Providers
Local logs are the primary source for every provider. Vendor APIs (Cursor Usage, GitHub Billing) are used only as supplementary truth-up — so attribution keeps working offline and when an upstream API breaks.
Coverage matrix
| Provider | Status | Source | Latency |
|---|---|---|---|
| Cursor | Live | ||
| Claude Code | Live | ||
| Codex CLI | Live | ||
| Copilot CLI | Live | ||
| Copilot Chat (VS Code) | Live | ||
| Continue, Cline, Roo Code, Aider, Gemini CLI | Coming soon |
Cursor
Budi reads Cursor's local SQLite database for per-message token counts and model info. The Cursor Usage API supplements this with dollar-level reconciliation.
Full guideClaude Code
Claude Code writes JSONL transcripts to ~/.claude/projects/. Budi tails these files in real-time for session, token, and cost data.
Full guideCodex CLI
OpenAI's Codex CLI writes JSONL session logs. Budi tails these the same way it handles Claude Code transcripts — automatic discovery, real-time attribution.
Full guideCopilot CLI
GitHub Copilot's CLI mode writes local session state files. Budi reads these for token counts and model info.
Full guideCopilot Chat (VS Code)
VS Code's Copilot Chat stores local session data that budi tails for output-only token counts. For dollar-accurate reconciliation, add a GitHub PAT scoped to billing.
Full guideContinue, Cline, Roo Code, Aider, Gemini CLI
Planned for 8.5.0. Same provider-plugin pattern as Cursor and Copilot Chat — local logs primary, no proxy.
How local tailing works
Budi's daemon watches the transcript and log directories that each agent writes to. It reads
new lines as they appear — like tail -f but structured. The daemon never injects itself into the request path between your editor and
the AI provider. This means:
- No added latency — the daemon is a passive reader.
- No proxy or gateway — nothing sits between your editor and the AI model.
- Works offline — local logs don't depend on network access.
- Survives API outages — if the vendor API is down, local tailing still works.
Copilot Chat and GitHub PATs
Without a PAT, on May-2026+ VS Code builds you get output-only token dollars — useful as a relative trend, but typically a fraction of the actual bill. With a PAT scoped to billing, budi reconciles to GitHub's exact dollar number per (date, model) bucket. Full PAT setup guide →
Org-managed licences see output-only forever — the user-level Billing API endpoint is empty for them. This is a GitHub limitation, not a budi one.