v0.9.15

What's New

Highlights since v0.9.14:

Added (CLI)

• codeburn context. See what fills a session's context window, by role, block type, and tool: an interactive terminal browser over your Claude Code and Codex sessions, separating the live window from compacted history and anchoring estimates to the exact API-reported context size. Also a Context page in the browser dashboard, and scriptable via codeburn context <id> --json. (#592)
• Zed provider. Zed's built-in agent is now tracked: per-request token usage with full cache fields, topped up to each thread's exact cumulative counter and validated token-for-token against a real store. (#594, format documented by @chatzinikolakisk in #480)
• codeburn audit. Per provider-and-model table of where every number comes from: calls, input, output, reasoning, cache read/write, cost. (#578)
• User price overrides for any model via codeburn price-override. (#560, thanks @ozymandiashh)
• open-design provider for per-model usage tracking. (#559, thanks @ozymandiashh)
• Browser dashboard is now fully mobile responsive (#582, thanks @ele-yufo; #589) and paints instantly with the local payload inlined, defaulting to today with fast-fail for offline paired devices. (#573)

Fixed (CLI)

• Cursor tokens are now Cursor's own numbers. Input comes from the per-conversation context meter instead of text-length guesses, credited once per conversation on a stable anchor so daily history stays consistent across re-scans; tools and shell commands come from the agent stream; Composer house models price at Cursor's published rates; figures are flagged estimated where they are. Fixes the class of report in #574 and closes #326. (#575)
• Copilot Chat users no longer see $0.00: VS Code core chatSessions journals are read for token counts. (#563, thanks @ozymandiashh)
• Codex sessions up to 4GB are parsed (streaming cap raised). (#569)
• Devin: ATIF v1.7 support (#570, thanks @tvcsantos) and friendly GPT model names with effort tiers. (#585)
• OpenCode skills and subagents breakdowns populated. (#557, thanks @KevNev19)
• Pi native skill loads classify as Skill, not Read. (#590)
• Cache read/write scoped to the selected period in web and devices CLI. (#586, thanks @ozymandiashh)
• Invalid dashboard periods return an error instead of exiting the web server. (#554, thanks @ozymandiashh)
• LiteLLM pricing snapshot refreshed; MiniMax-M3 follows MiniMax's new tiered pricing (standard tier $0.30/$1.20 per M). Daily cache bumped to v10 so history re-hydrates under the new Cursor accounting and pricing.

macOS menubar

• Local/Combined usage toggle backed by combined multi-device data in menubar-json. (#567, #568, thanks @ozymandiashh)
• The update dialog now detects a codeburn CLI too old to install menubar updates and shows the exact CLI upgrade command first. (#593)

Thanks to @ozymandiashh, @tvcsantos, @KevNev19, @ele-yufo, and @chatzinikolakisk for contributions this release.

The macOS menubar app is published separately as mac-v0.9.15.

What's Changed

• docs(changelog): credit @vaibhavarora14 for #486 by @iamtoruk in #552
• Raise streaming session cap to 4GB so large Codex sessions are counted by @iamtoruk in #569
• docs(readme): lead with a one-line definition for SEO/AI-answer consistency by @iamtoruk in #571
• fix(devin): add missing support for ATIF v1.7 by @tvcsantos in #570
• feat(pricing): add user price overrides for models by @ozymandiashh in #560
• feat(providers): add open-design provider by @ozymandiashh in #559
• fix(web): reject invalid dashboard periods without exiting by @ozymandiashh in #554
• Populate OpenCode skills and subagents breakdowns by @KevNev19 in #557
• fix(copilot): read VS Code chatSessions so Copilot Chat users see cost (#555) by @ozymandiashh in #563
• feat(menubar): expose combined multi-device usage in menubar-json (#566) by @ozymandiashh in #567
• feat(menubar): Local/Combined usage toggle — Swift, stacked on #567 (#566) by @ozymandiashh in #568
• perf(web): instant dashboard load, default to today, fast-fail offline peers by @iamtoruk in #573
• Clarify CodeBurn's open-source project status by @iamtoruk in #580
• feat: add codeburn audit token-source view by @iamtoruk in #578
• feat(web): mobile responsive web dashboard by @ele-yufo in #582
• Complete mobile responsiveness (header overflow + a11y) by @iamtoruk in #589
• fix: scope cache read/write to the selected period (web + devices CLI) (#583) by @ozymandiashh in #586
• fix(pi): classify native skill loads as Skill, not Read (#588) by @ozymandiashh in #590
• fix(devin): report friendly GPT model names with effort tier by @ozymandiashh in #585
• fix(cursor): use Cursor's real context tokens for input (#574) by @iamtoruk in #575
• feat(context): per-session context token tree across CLI, TUI, and dashboard by @iamtoruk in #592
• fix(menubar): refuse menubar updates from CLIs older than 0.9.9 by @iamtoruk in #593
• feat(providers): add Zed agent provider by @iamtoruk in #594
• release: 0.9.15 by @iamtoruk in #595

New Contributors

• @KevNev19 made their first contribution in #557
• @ele-yufo made their first contribution in #582

Full Changelog: v0.9.14...v0.9.15