feat(usage): add Cursor to the Usage page

[GimpyHand]

Summary

• Adds Cursor as a third Usage provider (alongside Claude and Codex) on web and mobile.
• Cursor agent transcripts have no token counts, so the server loads the same dashboard CSV export Cursor already exposes when desktop is signed in on the host (state.vscdb session → export-usage-events-csv), with a short on-disk cache and soft-fail so Claude/Codex still render if Cursor is missing.
• Prices Cursor export slugs at API-equivalent rates: Cursor-published Auto Cost / Composer / Grok rates for product models, LiteLLM after stripping effort/thinking suffixes for third-party names, and provider-reported CSV dollars when present. Bumps usage contract to v4.

Why this path

Claude/Codex usage comes from local JSONL ledgers. Cursor does not write token totals locally, so a pure filesystem walk cannot cover it. Reusing the signed-in desktop session to fetch the user's own usage export matches the existing host-trust model (the environment machine already holds provider identity) without inventing a parallel billing system.

Costs remain estimates, not subscription invoices. Auto is priced as Auto Cost because the export does not say which Auto mode ran.

Test plan

• vp test run apps/server/src/usage/usageCursorExport.test.ts apps/server/src/usage/usagePricing.test.ts apps/server/src/usage/usageAggregation.test.ts packages/shared/src/usageMerge.test.ts
• On a machine with Cursor desktop signed in, open Usage and confirm a Cursor series with non-zero tokens
• Confirm Included rows still contribute tokens and get API-equivalent dollars (not stuck at $0)
• On a machine without Cursor login, Usage still shows Claude/Codex and marks Cursor uncovered
• Multi-environment merge does not double-count the same Cursor export fingerprint

---

Note

Medium Risk
Reads local Cursor auth from SQLite and calls Cursor’s API with the desktop session token; cost figures remain estimates and merge logic changed for shared Cursor accounts.

Overview
Adds Cursor as a third Usage provider on web and mobile (contract v4), alongside Claude and Codex.

Because Cursor transcripts have no token counts, the server loads usage from Cursor’s dashboard CSV export when desktop is signed in: it reads cursorAuth/accessToken from state.vscdb, calls the export API with a session cookie, caches the CSV per account (~5 min freshness), and parses v1/v2/v3 CSV shapes into the same UsageRecord pipeline. Failures are soft so Claude/Codex still render; merge uses a account-scoped fingerprint (cursor-account + cursor-export:user_*) so the same login is not double-counted across machines or worktrees, with richest in-window data winning ties.

Pricing gains Cursor-published rates for Auto / Composer / Grok, strips effort/thinking suffixes for LiteLLM lookup, drops zero-cost LiteLLM auto rows, and leaves “Included” CSV costs unreported so API-equivalent pricing still applies.

UI surfaces Cursor in charts/tables and shows coverage notices only for Cursor missing/failed exports (isCursorCoverageGap), not unused Claude/Codex homes. User docs describe sources and Cursor coverage requirements.

^{Reviewed by Cursor Bugbot for commit 37535e3. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Add Cursor as a usage provider on the Usage page

• Adds a new usageCursorExport module that locates Cursor's local SQLite DB, extracts an auth token, fetches the dashboard CSV export, caches it with a TTL, and parses multiple CSV schema variants (v1/v2/v3) into UsageRecord arrays.
• Integrates Cursor records into UsageService.readSummary, soft-failing on errors and reporting a UsageSource with status and distinct session counts.
• Extends usagePricing with Cursor-native rate tables for Auto, Composer 2/2.5, and Grok 4.5 slugs; fixes zero-cost LiteLLM rows from masking these rates.
• Updates usageMerge so the richest data owner wins when multiple environments share a Cursor account fingerprint, and adds isCursorCoverageGap for surfacing missing/failed Cursor exports in UI notices.
• Adds cursor to provider ordering, labels, colors, and icons across both web and mobile Usage UIs.

^{Macroscope summarized 37535e3.}

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 5febdc74-5d3a-41d5-b8c8-0dfe2c960759

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[macroscopeapp]

One Effect-service convention issue found: loadCursorUsageRecords takes an HttpClient service instance as a parameter instead of acquiring it from the Effect environment, which erases the requirement from its Effect type. Details inline.

^{Posted via Macroscope — Effect Service Conventions}

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit:9cb738ead2

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[macroscopeapp]

Approvability

Verdict: Needs human review

This PR adds Cursor as a new usage provider, introducing external API integration to cursor.com, local credential extraction from SQLite databases, and new pricing/parsing logic. The scope of this new feature warrants human review.

^{You can customize Macroscope's approvability policy. Learn more.}

[macroscopeapp]

Two UI consistency findings in the changed usage web files. Both are small, localized fixes.

^{Posted via Macroscope — UI Consistency}

[cursor]

Cursor Bugbot has reviewed your changes using high effort and found 2 potential issues.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

Want fixes drafted automatically? Bugbot Autofix can create code changes for findings. A team admin can enable Autofix in the Cursor dashboard.

^{Reviewed by Cursor Bugbot for commit 37535e3. Configure here.}

[cursor]

Tokenless billed rows dropped

Medium Severity

parseCursorUsageCsv skips every row whose token columns are all zero before recording reportedCostUsd. Cursor exports often bill tooling models such as premium and agent_review with a numeric Cost and no tokens, so that spend never reaches the aggregator and Usage undercounts Cursor dollars.

^{Reviewed by Cursor Bugbot for commit 37535e3. Configure here.}

[cursor]

Auth DB probe short-circuits

Low Severity

findCursorStateDb returns the first existing state.vscdb candidate without checking for a usable cursorAuth/accessToken. A stale or logged-out DB earlier in the probe list (for example Windows Roaming ahead of LOCALAPPDATA) blocks later valid paths, so readLocalCursorExportAuth reports missing and Cursor stays uncovered.

Additional Locations (1)

• apps/server/src/usage/usageCursorExport.ts#L149-L156

^{Reviewed by Cursor Bugbot for commit 37535e3. Configure here.}

[t3dotgg]

Note

🤖 GPT-5.6 Sol responding on behalf of Theo

Closing this PR after an automated pass over open pull requests. Duplicates another usage-reporting proposal while adding a broad client, contract, or provider integration.

[GimpyHand]

Summary

• Adds Cursor as a third Usage provider (alongside Claude and Codex) on web and mobile.
• Cursor agent transcripts have no token counts, so the server loads the same dashboard CSV export Cursor already exposes when desktop is signed in on the host (state.vscdb session → export-usage-events-csv), with a short on-disk cache and soft-fail so Claude/Codex still render if Cursor is missing.
• Prices Cursor export slugs at API-equivalent rates: Cursor-published Auto Cost / Composer / Grok rates for product models, LiteLLM after stripping effort/thinking suffixes for third-party names, and provider-reported CSV dollars when present. Bumps usage contract to v4.

Why this path

Claude/Codex usage comes from local JSONL ledgers. Cursor does not write token totals locally, so a pure filesystem walk cannot cover it. Reusing the signed-in desktop session to fetch the user's own usage export matches the existing host-trust model (the environment machine already holds provider identity) without inventing a parallel billing system.

Costs remain estimates, not subscription invoices. Auto is priced as Auto Cost because the export does not say which Auto mode ran.

Test plan

• vp test run apps/server/src/usage/usageCursorExport.test.ts apps/server/src/usage/usagePricing.test.ts apps/server/src/usage/usageAggregation.test.ts packages/shared/src/usageMerge.test.ts
• On a machine with Cursor desktop signed in, open Usage and confirm a Cursor series with non-zero tokens
• Confirm Included rows still contribute tokens and get API-equivalent dollars (not stuck at $0)
• On a machine without Cursor login, Usage still shows Claude/Codex and marks Cursor uncovered
• Multi-environment merge does not double-count the same Cursor export fingerprint

---

Note

Medium Risk
Reads local Cursor auth from SQLite and calls Cursor’s API with the desktop session token; cost figures remain estimates and merge logic changed for shared Cursor accounts.

Overview
Adds Cursor as a third Usage provider on web and mobile (contract v4), alongside Claude and Codex.

Because Cursor transcripts have no token counts, the server loads usage from Cursor’s dashboard CSV export when desktop is signed in: it reads cursorAuth/accessToken from state.vscdb, calls the export API with a session cookie, caches the CSV per account (~5 min freshness), and parses v1/v2/v3 CSV shapes into the same UsageRecord pipeline. Failures are soft so Claude/Codex still render; merge uses a account-scoped fingerprint (cursor-account + cursor-export:user_*) so the same login is not double-counted across machines or worktrees, with richest in-window data winning ties.

Pricing gains Cursor-published rates for Auto / Composer / Grok, strips effort/thinking suffixes for LiteLLM lookup, drops zero-cost LiteLLM auto rows, and leaves “Included” CSV costs unreported so API-equivalent pricing still applies.

UI surfaces Cursor in charts/tables and shows coverage notices only for Cursor missing/failed exports (isCursorCoverageGap), not unused Claude/Codex homes. User docs describe sources and Cursor coverage requirements.

^{Reviewed by Cursor Bugbot for commit 37535e3. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Add Cursor as a usage provider on the Usage page

• Adds a new usageCursorExport module that locates Cursor's local SQLite DB, extracts an auth token, fetches the dashboard CSV export, caches it with a TTL, and parses multiple CSV schema variants (v1/v2/v3) into UsageRecord arrays.
• Integrates Cursor records into UsageService.readSummary, soft-failing on errors and reporting a UsageSource with status and distinct session counts.
• Extends usagePricing with Cursor-native rate tables for Auto, Composer 2/2.5, and Grok 4.5 slugs; fixes zero-cost LiteLLM rows from masking these rates.
• Updates usageMerge so the richest data owner wins when multiple environments share a Cursor account fingerprint, and adds isCursorCoverageGap for surfacing missing/failed Cursor exports in UI notices.
• Adds cursor to provider ordering, labels, colors, and icons across both web and mobile Usage UIs.

^{Macroscope summarized 37535e3.}

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 5febdc74-5d3a-41d5-b8c8-0dfe2c960759

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[macroscopeapp]

One Effect-service convention issue found: loadCursorUsageRecords takes an HttpClient service instance as a parameter instead of acquiring it from the Effect environment, which erases the requirement from its Effect type. Details inline.

^{Posted via Macroscope — Effect Service Conventions}

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit:9cb738ead2

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[macroscopeapp]

Approvability

Verdict: Needs human review

This PR adds Cursor as a new usage provider, introducing external API integration to cursor.com, local credential extraction from SQLite databases, and new pricing/parsing logic. The scope of this new feature warrants human review.

^{You can customize Macroscope's approvability policy. Learn more.}

[macroscopeapp]

Two UI consistency findings in the changed usage web files. Both are small, localized fixes.

^{Posted via Macroscope — UI Consistency}

[cursor]

Cursor Bugbot has reviewed your changes using high effort and found 2 potential issues.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

Want fixes drafted automatically? Bugbot Autofix can create code changes for findings. A team admin can enable Autofix in the Cursor dashboard.

^{Reviewed by Cursor Bugbot for commit 37535e3. Configure here.}

[cursor]

Tokenless billed rows dropped

Medium Severity

parseCursorUsageCsv skips every row whose token columns are all zero before recording reportedCostUsd. Cursor exports often bill tooling models such as premium and agent_review with a numeric Cost and no tokens, so that spend never reaches the aggregator and Usage undercounts Cursor dollars.

^{Reviewed by Cursor Bugbot for commit 37535e3. Configure here.}

[cursor]

Auth DB probe short-circuits

Low Severity

findCursorStateDb returns the first existing state.vscdb candidate without checking for a usable cursorAuth/accessToken. A stale or logged-out DB earlier in the probe list (for example Windows Roaming ahead of LOCALAPPDATA) blocks later valid paths, so readLocalCursorExportAuth reports missing and Cursor stays uncovered.

Additional Locations (1)

• apps/server/src/usage/usageCursorExport.ts#L149-L156

^{Reviewed by Cursor Bugbot for commit 37535e3. Configure here.}

[t3dotgg]

Note

🤖 GPT-5.6 Sol responding on behalf of Theo

Closing this PR after an automated pass over open pull requests. Duplicates another usage-reporting proposal while adding a broad client, contract, or provider integration.

[GimpyHand]

Summary

• Adds Cursor as a third Usage provider (alongside Claude and Codex) on web and mobile.
• Cursor agent transcripts have no token counts, so the server loads the same dashboard CSV export Cursor already exposes when desktop is signed in on the host (state.vscdb session → export-usage-events-csv), with a short on-disk cache and soft-fail so Claude/Codex still render if Cursor is missing.
• Prices Cursor export slugs at API-equivalent rates: Cursor-published Auto Cost / Composer / Grok rates for product models, LiteLLM after stripping effort/thinking suffixes for third-party names, and provider-reported CSV dollars when present. Bumps usage contract to v4.

Why this path

Claude/Codex usage comes from local JSONL ledgers. Cursor does not write token totals locally, so a pure filesystem walk cannot cover it. Reusing the signed-in desktop session to fetch the user's own usage export matches the existing host-trust model (the environment machine already holds provider identity) without inventing a parallel billing system.

Costs remain estimates, not subscription invoices. Auto is priced as Auto Cost because the export does not say which Auto mode ran.

Test plan

• vp test run apps/server/src/usage/usageCursorExport.test.ts apps/server/src/usage/usagePricing.test.ts apps/server/src/usage/usageAggregation.test.ts packages/shared/src/usageMerge.test.ts
• On a machine with Cursor desktop signed in, open Usage and confirm a Cursor series with non-zero tokens
• Confirm Included rows still contribute tokens and get API-equivalent dollars (not stuck at $0)
• On a machine without Cursor login, Usage still shows Claude/Codex and marks Cursor uncovered
• Multi-environment merge does not double-count the same Cursor export fingerprint

---

Note

Medium Risk
Reads local Cursor auth from SQLite and calls Cursor’s API with the desktop session token; cost figures remain estimates and merge logic changed for shared Cursor accounts.

Overview
Adds Cursor as a third Usage provider on web and mobile (contract v4), alongside Claude and Codex.

Because Cursor transcripts have no token counts, the server loads usage from Cursor’s dashboard CSV export when desktop is signed in: it reads cursorAuth/accessToken from state.vscdb, calls the export API with a session cookie, caches the CSV per account (~5 min freshness), and parses v1/v2/v3 CSV shapes into the same UsageRecord pipeline. Failures are soft so Claude/Codex still render; merge uses a account-scoped fingerprint (cursor-account + cursor-export:user_*) so the same login is not double-counted across machines or worktrees, with richest in-window data winning ties.

Pricing gains Cursor-published rates for Auto / Composer / Grok, strips effort/thinking suffixes for LiteLLM lookup, drops zero-cost LiteLLM auto rows, and leaves “Included” CSV costs unreported so API-equivalent pricing still applies.

UI surfaces Cursor in charts/tables and shows coverage notices only for Cursor missing/failed exports (isCursorCoverageGap), not unused Claude/Codex homes. User docs describe sources and Cursor coverage requirements.

^{Reviewed by Cursor Bugbot for commit 37535e3. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Add Cursor as a usage provider on the Usage page

• Adds a new usageCursorExport module that locates Cursor's local SQLite DB, extracts an auth token, fetches the dashboard CSV export, caches it with a TTL, and parses multiple CSV schema variants (v1/v2/v3) into UsageRecord arrays.
• Integrates Cursor records into UsageService.readSummary, soft-failing on errors and reporting a UsageSource with status and distinct session counts.
• Extends usagePricing with Cursor-native rate tables for Auto, Composer 2/2.5, and Grok 4.5 slugs; fixes zero-cost LiteLLM rows from masking these rates.
• Updates usageMerge so the richest data owner wins when multiple environments share a Cursor account fingerprint, and adds isCursorCoverageGap for surfacing missing/failed Cursor exports in UI notices.
• Adds cursor to provider ordering, labels, colors, and icons across both web and mobile Usage UIs.

^{Macroscope summarized 37535e3.}

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 5febdc74-5d3a-41d5-b8c8-0dfe2c960759

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[macroscopeapp]

One Effect-service convention issue found: loadCursorUsageRecords takes an HttpClient service instance as a parameter instead of acquiring it from the Effect environment, which erases the requirement from its Effect type. Details inline.

^{Posted via Macroscope — Effect Service Conventions}

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit:9cb738ead2

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[macroscopeapp]

Approvability

Verdict: Needs human review

This PR adds Cursor as a new usage provider, introducing external API integration to cursor.com, local credential extraction from SQLite databases, and new pricing/parsing logic. The scope of this new feature warrants human review.

^{You can customize Macroscope's approvability policy. Learn more.}

[macroscopeapp]

Two UI consistency findings in the changed usage web files. Both are small, localized fixes.

^{Posted via Macroscope — UI Consistency}

[cursor]

Cursor Bugbot has reviewed your changes using high effort and found 2 potential issues.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

Want fixes drafted automatically? Bugbot Autofix can create code changes for findings. A team admin can enable Autofix in the Cursor dashboard.

^{Reviewed by Cursor Bugbot for commit 37535e3. Configure here.}

[cursor]

Tokenless billed rows dropped

Medium Severity

parseCursorUsageCsv skips every row whose token columns are all zero before recording reportedCostUsd. Cursor exports often bill tooling models such as premium and agent_review with a numeric Cost and no tokens, so that spend never reaches the aggregator and Usage undercounts Cursor dollars.

^{Reviewed by Cursor Bugbot for commit 37535e3. Configure here.}

[cursor]

Auth DB probe short-circuits

Low Severity

findCursorStateDb returns the first existing state.vscdb candidate without checking for a usable cursorAuth/accessToken. A stale or logged-out DB earlier in the probe list (for example Windows Roaming ahead of LOCALAPPDATA) blocks later valid paths, so readLocalCursorExportAuth reports missing and Cursor stays uncovered.

Additional Locations (1)

• apps/server/src/usage/usageCursorExport.ts#L149-L156

^{Reviewed by Cursor Bugbot for commit 37535e3. Configure here.}

[t3dotgg]

Note

🤖 GPT-5.6 Sol responding on behalf of Theo

Closing this PR after an automated pass over open pull requests. Duplicates another usage-reporting proposal while adding a broad client, contract, or provider integration.

[GimpyHand]

Summary

• Adds Cursor as a third Usage provider (alongside Claude and Codex) on web and mobile.
• Cursor agent transcripts have no token counts, so the server loads the same dashboard CSV export Cursor already exposes when desktop is signed in on the host (state.vscdb session → export-usage-events-csv), with a short on-disk cache and soft-fail so Claude/Codex still render if Cursor is missing.
• Prices Cursor export slugs at API-equivalent rates: Cursor-published Auto Cost / Composer / Grok rates for product models, LiteLLM after stripping effort/thinking suffixes for third-party names, and provider-reported CSV dollars when present. Bumps usage contract to v4.

Why this path

Claude/Codex usage comes from local JSONL ledgers. Cursor does not write token totals locally, so a pure filesystem walk cannot cover it. Reusing the signed-in desktop session to fetch the user's own usage export matches the existing host-trust model (the environment machine already holds provider identity) without inventing a parallel billing system.

Costs remain estimates, not subscription invoices. Auto is priced as Auto Cost because the export does not say which Auto mode ran.

Test plan

• vp test run apps/server/src/usage/usageCursorExport.test.ts apps/server/src/usage/usagePricing.test.ts apps/server/src/usage/usageAggregation.test.ts packages/shared/src/usageMerge.test.ts
• On a machine with Cursor desktop signed in, open Usage and confirm a Cursor series with non-zero tokens
• Confirm Included rows still contribute tokens and get API-equivalent dollars (not stuck at $0)
• On a machine without Cursor login, Usage still shows Claude/Codex and marks Cursor uncovered
• Multi-environment merge does not double-count the same Cursor export fingerprint

---

Note

Medium Risk
Reads local Cursor auth from SQLite and calls Cursor’s API with the desktop session token; cost figures remain estimates and merge logic changed for shared Cursor accounts.

Overview
Adds Cursor as a third Usage provider on web and mobile (contract v4), alongside Claude and Codex.

Because Cursor transcripts have no token counts, the server loads usage from Cursor’s dashboard CSV export when desktop is signed in: it reads cursorAuth/accessToken from state.vscdb, calls the export API with a session cookie, caches the CSV per account (~5 min freshness), and parses v1/v2/v3 CSV shapes into the same UsageRecord pipeline. Failures are soft so Claude/Codex still render; merge uses a account-scoped fingerprint (cursor-account + cursor-export:user_*) so the same login is not double-counted across machines or worktrees, with richest in-window data winning ties.

Pricing gains Cursor-published rates for Auto / Composer / Grok, strips effort/thinking suffixes for LiteLLM lookup, drops zero-cost LiteLLM auto rows, and leaves “Included” CSV costs unreported so API-equivalent pricing still applies.

UI surfaces Cursor in charts/tables and shows coverage notices only for Cursor missing/failed exports (isCursorCoverageGap), not unused Claude/Codex homes. User docs describe sources and Cursor coverage requirements.

^{Reviewed by Cursor Bugbot for commit 37535e3. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Add Cursor as a usage provider on the Usage page

• Adds a new usageCursorExport module that locates Cursor's local SQLite DB, extracts an auth token, fetches the dashboard CSV export, caches it with a TTL, and parses multiple CSV schema variants (v1/v2/v3) into UsageRecord arrays.
• Integrates Cursor records into UsageService.readSummary, soft-failing on errors and reporting a UsageSource with status and distinct session counts.
• Extends usagePricing with Cursor-native rate tables for Auto, Composer 2/2.5, and Grok 4.5 slugs; fixes zero-cost LiteLLM rows from masking these rates.
• Updates usageMerge so the richest data owner wins when multiple environments share a Cursor account fingerprint, and adds isCursorCoverageGap for surfacing missing/failed Cursor exports in UI notices.
• Adds cursor to provider ordering, labels, colors, and icons across both web and mobile Usage UIs.

^{Macroscope summarized 37535e3.}

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 5febdc74-5d3a-41d5-b8c8-0dfe2c960759

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[macroscopeapp]

One Effect-service convention issue found: loadCursorUsageRecords takes an HttpClient service instance as a parameter instead of acquiring it from the Effect environment, which erases the requirement from its Effect type. Details inline.

^{Posted via Macroscope — Effect Service Conventions}

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit:9cb738ead2

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[macroscopeapp]

Approvability

Verdict: Needs human review

This PR adds Cursor as a new usage provider, introducing external API integration to cursor.com, local credential extraction from SQLite databases, and new pricing/parsing logic. The scope of this new feature warrants human review.

^{You can customize Macroscope's approvability policy. Learn more.}

[macroscopeapp]

Two UI consistency findings in the changed usage web files. Both are small, localized fixes.

^{Posted via Macroscope — UI Consistency}

[cursor]

Cursor Bugbot has reviewed your changes using high effort and found 2 potential issues.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

Want fixes drafted automatically? Bugbot Autofix can create code changes for findings. A team admin can enable Autofix in the Cursor dashboard.

^{Reviewed by Cursor Bugbot for commit 37535e3. Configure here.}

[cursor]

Tokenless billed rows dropped

Medium Severity

parseCursorUsageCsv skips every row whose token columns are all zero before recording reportedCostUsd. Cursor exports often bill tooling models such as premium and agent_review with a numeric Cost and no tokens, so that spend never reaches the aggregator and Usage undercounts Cursor dollars.

^{Reviewed by Cursor Bugbot for commit 37535e3. Configure here.}

[cursor]

Auth DB probe short-circuits

Low Severity

findCursorStateDb returns the first existing state.vscdb candidate without checking for a usable cursorAuth/accessToken. A stale or logged-out DB earlier in the probe list (for example Windows Roaming ahead of LOCALAPPDATA) blocks later valid paths, so readLocalCursorExportAuth reports missing and Cursor stays uncovered.

Additional Locations (1)

• apps/server/src/usage/usageCursorExport.ts#L149-L156

^{Reviewed by Cursor Bugbot for commit 37535e3. Configure here.}

[t3dotgg]

Note

🤖 GPT-5.6 Sol responding on behalf of Theo

Closing this PR after an automated pass over open pull requests. Duplicates another usage-reporting proposal while adding a broad client, contract, or provider integration.

[GimpyHand]

Summary

• Adds Cursor as a third Usage provider (alongside Claude and Codex) on web and mobile.
• Cursor agent transcripts have no token counts, so the server loads the same dashboard CSV export Cursor already exposes when desktop is signed in on the host (state.vscdb session → export-usage-events-csv), with a short on-disk cache and soft-fail so Claude/Codex still render if Cursor is missing.
• Prices Cursor export slugs at API-equivalent rates: Cursor-published Auto Cost / Composer / Grok rates for product models, LiteLLM after stripping effort/thinking suffixes for third-party names, and provider-reported CSV dollars when present. Bumps usage contract to v4.

Why this path

Claude/Codex usage comes from local JSONL ledgers. Cursor does not write token totals locally, so a pure filesystem walk cannot cover it. Reusing the signed-in desktop session to fetch the user's own usage export matches the existing host-trust model (the environment machine already holds provider identity) without inventing a parallel billing system.

Costs remain estimates, not subscription invoices. Auto is priced as Auto Cost because the export does not say which Auto mode ran.

Test plan

• vp test run apps/server/src/usage/usageCursorExport.test.ts apps/server/src/usage/usagePricing.test.ts apps/server/src/usage/usageAggregation.test.ts packages/shared/src/usageMerge.test.ts
• On a machine with Cursor desktop signed in, open Usage and confirm a Cursor series with non-zero tokens
• Confirm Included rows still contribute tokens and get API-equivalent dollars (not stuck at $0)
• On a machine without Cursor login, Usage still shows Claude/Codex and marks Cursor uncovered
• Multi-environment merge does not double-count the same Cursor export fingerprint

---

Note

Medium Risk
Reads local Cursor auth from SQLite and calls Cursor’s API with the desktop session token; cost figures remain estimates and merge logic changed for shared Cursor accounts.

Overview
Adds Cursor as a third Usage provider on web and mobile (contract v4), alongside Claude and Codex.

Because Cursor transcripts have no token counts, the server loads usage from Cursor’s dashboard CSV export when desktop is signed in: it reads cursorAuth/accessToken from state.vscdb, calls the export API with a session cookie, caches the CSV per account (~5 min freshness), and parses v1/v2/v3 CSV shapes into the same UsageRecord pipeline. Failures are soft so Claude/Codex still render; merge uses a account-scoped fingerprint (cursor-account + cursor-export:user_*) so the same login is not double-counted across machines or worktrees, with richest in-window data winning ties.

Pricing gains Cursor-published rates for Auto / Composer / Grok, strips effort/thinking suffixes for LiteLLM lookup, drops zero-cost LiteLLM auto rows, and leaves “Included” CSV costs unreported so API-equivalent pricing still applies.

UI surfaces Cursor in charts/tables and shows coverage notices only for Cursor missing/failed exports (isCursorCoverageGap), not unused Claude/Codex homes. User docs describe sources and Cursor coverage requirements.

^{Reviewed by Cursor Bugbot for commit 37535e3. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Add Cursor as a usage provider on the Usage page

• Adds a new usageCursorExport module that locates Cursor's local SQLite DB, extracts an auth token, fetches the dashboard CSV export, caches it with a TTL, and parses multiple CSV schema variants (v1/v2/v3) into UsageRecord arrays.
• Integrates Cursor records into UsageService.readSummary, soft-failing on errors and reporting a UsageSource with status and distinct session counts.
• Extends usagePricing with Cursor-native rate tables for Auto, Composer 2/2.5, and Grok 4.5 slugs; fixes zero-cost LiteLLM rows from masking these rates.
• Updates usageMerge so the richest data owner wins when multiple environments share a Cursor account fingerprint, and adds isCursorCoverageGap for surfacing missing/failed Cursor exports in UI notices.
• Adds cursor to provider ordering, labels, colors, and icons across both web and mobile Usage UIs.

^{Macroscope summarized 37535e3.}

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 5febdc74-5d3a-41d5-b8c8-0dfe2c960759

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[macroscopeapp]

One Effect-service convention issue found: loadCursorUsageRecords takes an HttpClient service instance as a parameter instead of acquiring it from the Effect environment, which erases the requirement from its Effect type. Details inline.

^{Posted via Macroscope — Effect Service Conventions}

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit:9cb738ead2

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[macroscopeapp]

Approvability

Verdict: Needs human review

This PR adds Cursor as a new usage provider, introducing external API integration to cursor.com, local credential extraction from SQLite databases, and new pricing/parsing logic. The scope of this new feature warrants human review.

^{You can customize Macroscope's approvability policy. Learn more.}

[macroscopeapp]

Two UI consistency findings in the changed usage web files. Both are small, localized fixes.

^{Posted via Macroscope — UI Consistency}

[cursor]

Cursor Bugbot has reviewed your changes using high effort and found 2 potential issues.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

Want fixes drafted automatically? Bugbot Autofix can create code changes for findings. A team admin can enable Autofix in the Cursor dashboard.

^{Reviewed by Cursor Bugbot for commit 37535e3. Configure here.}

[cursor]

Tokenless billed rows dropped

Medium Severity

parseCursorUsageCsv skips every row whose token columns are all zero before recording reportedCostUsd. Cursor exports often bill tooling models such as premium and agent_review with a numeric Cost and no tokens, so that spend never reaches the aggregator and Usage undercounts Cursor dollars.

^{Reviewed by Cursor Bugbot for commit 37535e3. Configure here.}

[cursor]

Auth DB probe short-circuits

Low Severity

findCursorStateDb returns the first existing state.vscdb candidate without checking for a usable cursorAuth/accessToken. A stale or logged-out DB earlier in the probe list (for example Windows Roaming ahead of LOCALAPPDATA) blocks later valid paths, so readLocalCursorExportAuth reports missing and Cursor stays uncovered.

Additional Locations (1)

• apps/server/src/usage/usageCursorExport.ts#L149-L156

^{Reviewed by Cursor Bugbot for commit 37535e3. Configure here.}

[t3dotgg]

Note

🤖 GPT-5.6 Sol responding on behalf of Theo

Closing this PR after an automated pass over open pull requests. Duplicates another usage-reporting proposal while adding a broad client, contract, or provider integration.

[GimpyHand]

Summary

• Adds Cursor as a third Usage provider (alongside Claude and Codex) on web and mobile.
• Cursor agent transcripts have no token counts, so the server loads the same dashboard CSV export Cursor already exposes when desktop is signed in on the host (state.vscdb session → export-usage-events-csv), with a short on-disk cache and soft-fail so Claude/Codex still render if Cursor is missing.
• Prices Cursor export slugs at API-equivalent rates: Cursor-published Auto Cost / Composer / Grok rates for product models, LiteLLM after stripping effort/thinking suffixes for third-party names, and provider-reported CSV dollars when present. Bumps usage contract to v4.

Why this path

Claude/Codex usage comes from local JSONL ledgers. Cursor does not write token totals locally, so a pure filesystem walk cannot cover it. Reusing the signed-in desktop session to fetch the user's own usage export matches the existing host-trust model (the environment machine already holds provider identity) without inventing a parallel billing system.

Costs remain estimates, not subscription invoices. Auto is priced as Auto Cost because the export does not say which Auto mode ran.

Test plan

• vp test run apps/server/src/usage/usageCursorExport.test.ts apps/server/src/usage/usagePricing.test.ts apps/server/src/usage/usageAggregation.test.ts packages/shared/src/usageMerge.test.ts
• On a machine with Cursor desktop signed in, open Usage and confirm a Cursor series with non-zero tokens
• Confirm Included rows still contribute tokens and get API-equivalent dollars (not stuck at $0)
• On a machine without Cursor login, Usage still shows Claude/Codex and marks Cursor uncovered
• Multi-environment merge does not double-count the same Cursor export fingerprint

---

Note

Medium Risk
Reads local Cursor auth from SQLite and calls Cursor’s API with the desktop session token; cost figures remain estimates and merge logic changed for shared Cursor accounts.

Overview
Adds Cursor as a third Usage provider on web and mobile (contract v4), alongside Claude and Codex.

Because Cursor transcripts have no token counts, the server loads usage from Cursor’s dashboard CSV export when desktop is signed in: it reads cursorAuth/accessToken from state.vscdb, calls the export API with a session cookie, caches the CSV per account (~5 min freshness), and parses v1/v2/v3 CSV shapes into the same UsageRecord pipeline. Failures are soft so Claude/Codex still render; merge uses a account-scoped fingerprint (cursor-account + cursor-export:user_*) so the same login is not double-counted across machines or worktrees, with richest in-window data winning ties.

Pricing gains Cursor-published rates for Auto / Composer / Grok, strips effort/thinking suffixes for LiteLLM lookup, drops zero-cost LiteLLM auto rows, and leaves “Included” CSV costs unreported so API-equivalent pricing still applies.

UI surfaces Cursor in charts/tables and shows coverage notices only for Cursor missing/failed exports (isCursorCoverageGap), not unused Claude/Codex homes. User docs describe sources and Cursor coverage requirements.

^{Reviewed by Cursor Bugbot for commit 37535e3. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Add Cursor as a usage provider on the Usage page

• Adds a new usageCursorExport module that locates Cursor's local SQLite DB, extracts an auth token, fetches the dashboard CSV export, caches it with a TTL, and parses multiple CSV schema variants (v1/v2/v3) into UsageRecord arrays.
• Integrates Cursor records into UsageService.readSummary, soft-failing on errors and reporting a UsageSource with status and distinct session counts.
• Extends usagePricing with Cursor-native rate tables for Auto, Composer 2/2.5, and Grok 4.5 slugs; fixes zero-cost LiteLLM rows from masking these rates.
• Updates usageMerge so the richest data owner wins when multiple environments share a Cursor account fingerprint, and adds isCursorCoverageGap for surfacing missing/failed Cursor exports in UI notices.
• Adds cursor to provider ordering, labels, colors, and icons across both web and mobile Usage UIs.

^{Macroscope summarized 37535e3.}

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 5febdc74-5d3a-41d5-b8c8-0dfe2c960759

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[macroscopeapp]

One Effect-service convention issue found: loadCursorUsageRecords takes an HttpClient service instance as a parameter instead of acquiring it from the Effect environment, which erases the requirement from its Effect type. Details inline.

^{Posted via Macroscope — Effect Service Conventions}

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit:9cb738ead2

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[macroscopeapp]

Approvability

Verdict: Needs human review

This PR adds Cursor as a new usage provider, introducing external API integration to cursor.com, local credential extraction from SQLite databases, and new pricing/parsing logic. The scope of this new feature warrants human review.

^{You can customize Macroscope's approvability policy. Learn more.}

[macroscopeapp]

Two UI consistency findings in the changed usage web files. Both are small, localized fixes.

^{Posted via Macroscope — UI Consistency}

[cursor]

Cursor Bugbot has reviewed your changes using high effort and found 2 potential issues.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

Want fixes drafted automatically? Bugbot Autofix can create code changes for findings. A team admin can enable Autofix in the Cursor dashboard.

^{Reviewed by Cursor Bugbot for commit 37535e3. Configure here.}

[cursor]

Tokenless billed rows dropped

Medium Severity

parseCursorUsageCsv skips every row whose token columns are all zero before recording reportedCostUsd. Cursor exports often bill tooling models such as premium and agent_review with a numeric Cost and no tokens, so that spend never reaches the aggregator and Usage undercounts Cursor dollars.

^{Reviewed by Cursor Bugbot for commit 37535e3. Configure here.}

[cursor]

Auth DB probe short-circuits

Low Severity

findCursorStateDb returns the first existing state.vscdb candidate without checking for a usable cursorAuth/accessToken. A stale or logged-out DB earlier in the probe list (for example Windows Roaming ahead of LOCALAPPDATA) blocks later valid paths, so readLocalCursorExportAuth reports missing and Cursor stays uncovered.

Additional Locations (1)

• apps/server/src/usage/usageCursorExport.ts#L149-L156

^{Reviewed by Cursor Bugbot for commit 37535e3. Configure here.}

[t3dotgg]

Note

🤖 GPT-5.6 Sol responding on behalf of Theo

Closing this PR after an automated pass over open pull requests. Duplicates another usage-reporting proposal while adding a broad client, contract, or provider integration.

[GimpyHand]

Summary

• Adds Cursor as a third Usage provider (alongside Claude and Codex) on web and mobile.
• Cursor agent transcripts have no token counts, so the server loads the same dashboard CSV export Cursor already exposes when desktop is signed in on the host (state.vscdb session → export-usage-events-csv), with a short on-disk cache and soft-fail so Claude/Codex still render if Cursor is missing.
• Prices Cursor export slugs at API-equivalent rates: Cursor-published Auto Cost / Composer / Grok rates for product models, LiteLLM after stripping effort/thinking suffixes for third-party names, and provider-reported CSV dollars when present. Bumps usage contract to v4.

Why this path

Claude/Codex usage comes from local JSONL ledgers. Cursor does not write token totals locally, so a pure filesystem walk cannot cover it. Reusing the signed-in desktop session to fetch the user's own usage export matches the existing host-trust model (the environment machine already holds provider identity) without inventing a parallel billing system.

Costs remain estimates, not subscription invoices. Auto is priced as Auto Cost because the export does not say which Auto mode ran.

Test plan

• vp test run apps/server/src/usage/usageCursorExport.test.ts apps/server/src/usage/usagePricing.test.ts apps/server/src/usage/usageAggregation.test.ts packages/shared/src/usageMerge.test.ts
• On a machine with Cursor desktop signed in, open Usage and confirm a Cursor series with non-zero tokens
• Confirm Included rows still contribute tokens and get API-equivalent dollars (not stuck at $0)
• On a machine without Cursor login, Usage still shows Claude/Codex and marks Cursor uncovered
• Multi-environment merge does not double-count the same Cursor export fingerprint

---

Note

Medium Risk
Reads local Cursor auth from SQLite and calls Cursor’s API with the desktop session token; cost figures remain estimates and merge logic changed for shared Cursor accounts.

Overview
Adds Cursor as a third Usage provider on web and mobile (contract v4), alongside Claude and Codex.

Because Cursor transcripts have no token counts, the server loads usage from Cursor’s dashboard CSV export when desktop is signed in: it reads cursorAuth/accessToken from state.vscdb, calls the export API with a session cookie, caches the CSV per account (~5 min freshness), and parses v1/v2/v3 CSV shapes into the same UsageRecord pipeline. Failures are soft so Claude/Codex still render; merge uses a account-scoped fingerprint (cursor-account + cursor-export:user_*) so the same login is not double-counted across machines or worktrees, with richest in-window data winning ties.

Pricing gains Cursor-published rates for Auto / Composer / Grok, strips effort/thinking suffixes for LiteLLM lookup, drops zero-cost LiteLLM auto rows, and leaves “Included” CSV costs unreported so API-equivalent pricing still applies.

UI surfaces Cursor in charts/tables and shows coverage notices only for Cursor missing/failed exports (isCursorCoverageGap), not unused Claude/Codex homes. User docs describe sources and Cursor coverage requirements.

^{Reviewed by Cursor Bugbot for commit 37535e3. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Add Cursor as a usage provider on the Usage page

• Adds a new usageCursorExport module that locates Cursor's local SQLite DB, extracts an auth token, fetches the dashboard CSV export, caches it with a TTL, and parses multiple CSV schema variants (v1/v2/v3) into UsageRecord arrays.
• Integrates Cursor records into UsageService.readSummary, soft-failing on errors and reporting a UsageSource with status and distinct session counts.
• Extends usagePricing with Cursor-native rate tables for Auto, Composer 2/2.5, and Grok 4.5 slugs; fixes zero-cost LiteLLM rows from masking these rates.
• Updates usageMerge so the richest data owner wins when multiple environments share a Cursor account fingerprint, and adds isCursorCoverageGap for surfacing missing/failed Cursor exports in UI notices.
• Adds cursor to provider ordering, labels, colors, and icons across both web and mobile Usage UIs.

^{Macroscope summarized 37535e3.}

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 5febdc74-5d3a-41d5-b8c8-0dfe2c960759

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[macroscopeapp]

One Effect-service convention issue found: loadCursorUsageRecords takes an HttpClient service instance as a parameter instead of acquiring it from the Effect environment, which erases the requirement from its Effect type. Details inline.

^{Posted via Macroscope — Effect Service Conventions}

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit:9cb738ead2

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[macroscopeapp]

Approvability

Verdict: Needs human review

This PR adds Cursor as a new usage provider, introducing external API integration to cursor.com, local credential extraction from SQLite databases, and new pricing/parsing logic. The scope of this new feature warrants human review.

^{You can customize Macroscope's approvability policy. Learn more.}

[macroscopeapp]

Two UI consistency findings in the changed usage web files. Both are small, localized fixes.

^{Posted via Macroscope — UI Consistency}

[cursor]

Cursor Bugbot has reviewed your changes using high effort and found 2 potential issues.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

Want fixes drafted automatically? Bugbot Autofix can create code changes for findings. A team admin can enable Autofix in the Cursor dashboard.

^{Reviewed by Cursor Bugbot for commit 37535e3. Configure here.}

[cursor]

Tokenless billed rows dropped

Medium Severity

parseCursorUsageCsv skips every row whose token columns are all zero before recording reportedCostUsd. Cursor exports often bill tooling models such as premium and agent_review with a numeric Cost and no tokens, so that spend never reaches the aggregator and Usage undercounts Cursor dollars.

^{Reviewed by Cursor Bugbot for commit 37535e3. Configure here.}

[cursor]

Auth DB probe short-circuits

Low Severity

findCursorStateDb returns the first existing state.vscdb candidate without checking for a usable cursorAuth/accessToken. A stale or logged-out DB earlier in the probe list (for example Windows Roaming ahead of LOCALAPPDATA) blocks later valid paths, so readLocalCursorExportAuth reports missing and Cursor stays uncovered.

Additional Locations (1)

• apps/server/src/usage/usageCursorExport.ts#L149-L156

^{Reviewed by Cursor Bugbot for commit 37535e3. Configure here.}

[t3dotgg]

Note

🤖 GPT-5.6 Sol responding on behalf of Theo

Closing this PR after an automated pass over open pull requests. Duplicates another usage-reporting proposal while adding a broad client, contract, or provider integration.

[GimpyHand]

Summary

• Adds Cursor as a third Usage provider (alongside Claude and Codex) on web and mobile.
• Cursor agent transcripts have no token counts, so the server loads the same dashboard CSV export Cursor already exposes when desktop is signed in on the host (state.vscdb session → export-usage-events-csv), with a short on-disk cache and soft-fail so Claude/Codex still render if Cursor is missing.
• Prices Cursor export slugs at API-equivalent rates: Cursor-published Auto Cost / Composer / Grok rates for product models, LiteLLM after stripping effort/thinking suffixes for third-party names, and provider-reported CSV dollars when present. Bumps usage contract to v4.

Why this path

Claude/Codex usage comes from local JSONL ledgers. Cursor does not write token totals locally, so a pure filesystem walk cannot cover it. Reusing the signed-in desktop session to fetch the user's own usage export matches the existing host-trust model (the environment machine already holds provider identity) without inventing a parallel billing system.

Costs remain estimates, not subscription invoices. Auto is priced as Auto Cost because the export does not say which Auto mode ran.

Test plan

• vp test run apps/server/src/usage/usageCursorExport.test.ts apps/server/src/usage/usagePricing.test.ts apps/server/src/usage/usageAggregation.test.ts packages/shared/src/usageMerge.test.ts
• On a machine with Cursor desktop signed in, open Usage and confirm a Cursor series with non-zero tokens
• Confirm Included rows still contribute tokens and get API-equivalent dollars (not stuck at $0)
• On a machine without Cursor login, Usage still shows Claude/Codex and marks Cursor uncovered
• Multi-environment merge does not double-count the same Cursor export fingerprint

---

Note

Medium Risk
Reads local Cursor auth from SQLite and calls Cursor’s API with the desktop session token; cost figures remain estimates and merge logic changed for shared Cursor accounts.

Overview
Adds Cursor as a third Usage provider on web and mobile (contract v4), alongside Claude and Codex.

Because Cursor transcripts have no token counts, the server loads usage from Cursor’s dashboard CSV export when desktop is signed in: it reads cursorAuth/accessToken from state.vscdb, calls the export API with a session cookie, caches the CSV per account (~5 min freshness), and parses v1/v2/v3 CSV shapes into the same UsageRecord pipeline. Failures are soft so Claude/Codex still render; merge uses a account-scoped fingerprint (cursor-account + cursor-export:user_*) so the same login is not double-counted across machines or worktrees, with richest in-window data winning ties.

Pricing gains Cursor-published rates for Auto / Composer / Grok, strips effort/thinking suffixes for LiteLLM lookup, drops zero-cost LiteLLM auto rows, and leaves “Included” CSV costs unreported so API-equivalent pricing still applies.

UI surfaces Cursor in charts/tables and shows coverage notices only for Cursor missing/failed exports (isCursorCoverageGap), not unused Claude/Codex homes. User docs describe sources and Cursor coverage requirements.

^{Reviewed by Cursor Bugbot for commit 37535e3. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Add Cursor as a usage provider on the Usage page

• Adds a new usageCursorExport module that locates Cursor's local SQLite DB, extracts an auth token, fetches the dashboard CSV export, caches it with a TTL, and parses multiple CSV schema variants (v1/v2/v3) into UsageRecord arrays.
• Integrates Cursor records into UsageService.readSummary, soft-failing on errors and reporting a UsageSource with status and distinct session counts.
• Extends usagePricing with Cursor-native rate tables for Auto, Composer 2/2.5, and Grok 4.5 slugs; fixes zero-cost LiteLLM rows from masking these rates.
• Updates usageMerge so the richest data owner wins when multiple environments share a Cursor account fingerprint, and adds isCursorCoverageGap for surfacing missing/failed Cursor exports in UI notices.
• Adds cursor to provider ordering, labels, colors, and icons across both web and mobile Usage UIs.

^{Macroscope summarized 37535e3.}

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 5febdc74-5d3a-41d5-b8c8-0dfe2c960759

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[macroscopeapp]

One Effect-service convention issue found: loadCursorUsageRecords takes an HttpClient service instance as a parameter instead of acquiring it from the Effect environment, which erases the requirement from its Effect type. Details inline.

^{Posted via Macroscope — Effect Service Conventions}

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit:9cb738ead2

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[macroscopeapp]

Approvability

Verdict: Needs human review

This PR adds Cursor as a new usage provider, introducing external API integration to cursor.com, local credential extraction from SQLite databases, and new pricing/parsing logic. The scope of this new feature warrants human review.

^{You can customize Macroscope's approvability policy. Learn more.}

[macroscopeapp]

Two UI consistency findings in the changed usage web files. Both are small, localized fixes.

^{Posted via Macroscope — UI Consistency}

[cursor]

Cursor Bugbot has reviewed your changes using high effort and found 2 potential issues.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

Want fixes drafted automatically? Bugbot Autofix can create code changes for findings. A team admin can enable Autofix in the Cursor dashboard.

^{Reviewed by Cursor Bugbot for commit 37535e3. Configure here.}

[cursor]

Tokenless billed rows dropped

Medium Severity

parseCursorUsageCsv skips every row whose token columns are all zero before recording reportedCostUsd. Cursor exports often bill tooling models such as premium and agent_review with a numeric Cost and no tokens, so that spend never reaches the aggregator and Usage undercounts Cursor dollars.

^{Reviewed by Cursor Bugbot for commit 37535e3. Configure here.}

[cursor]

Auth DB probe short-circuits

Low Severity

findCursorStateDb returns the first existing state.vscdb candidate without checking for a usable cursorAuth/accessToken. A stale or logged-out DB earlier in the probe list (for example Windows Roaming ahead of LOCALAPPDATA) blocks later valid paths, so readLocalCursorExportAuth reports missing and Cursor stays uncovered.

Additional Locations (1)

• apps/server/src/usage/usageCursorExport.ts#L149-L156

^{Reviewed by Cursor Bugbot for commit 37535e3. Configure here.}

[t3dotgg]

Note

🤖 GPT-5.6 Sol responding on behalf of Theo

Closing this PR after an automated pass over open pull requests. Duplicates another usage-reporting proposal while adding a broad client, contract, or provider integration.