OpenUsage ships with provider integrations covering coding agents, API platforms, and local tools. All providers are auto-detected when available — no manual config needed.
Detection: claude binary + ~/.claude directory
Tracks daily activity, per-model token usage, 5-hour billing block computation, burn rate, and cost estimation.
Detection: cursor binary + local SQLite databases
Tracks plan spend and limits, per-model aggregation, Composer sessions, and AI code scoring. Uses a hybrid approach — API endpoints plus local SQLite DB reads.
Detection: gh CLI with Copilot extension installed
Tracks chat and completions quota, org billing, org metrics, and session tracking.
Detection: codex binary + ~/.codex directory
Tracks session tokens, per-model and per-client breakdown, credits, and rate limits.
Detection: gemini binary + ~/.gemini directory
Tracks OAuth status, conversation count, per-model tokens, and quota API data.
Detection: agy binary + ~/.gemini/antigravity-cli directory
Tracks status-line context usage, session tokens, model quotas, workspace, and agent state without reading credentials or estimating spend.
Detection: OPENCODE_API_KEY or ZEN_API_KEY environment variable
Tracks credits, activity, and generation stats via an OpenRouter-compatible API.
Detection: OLLAMA_HOST environment variable or ollama binary
Tracks local server models, per-model usage, and optional cloud billing.
Detection: OPENROUTER_API_KEY environment variable
Tracks credits, activity, generation stats, and per-model breakdown across multiple API endpoints.
Detection: OPENAI_API_KEY environment variable
Tracks rate limits via lightweight header probing.
Detection: ANTHROPIC_API_KEY environment variable
Tracks rate limits via lightweight header probing.
Detection: AZURE_OPENAI_API_KEY or AZURE_API_KEY, plus AZURE_RESOURCE_NAME (or AZURE_OPENAI_ENDPOINT) for the endpoint
Tracks rate limits via lightweight header probing against your Azure OpenAI resource endpoint. Shares OpenCode's AZURE_API_KEY / AZURE_RESOURCE_NAME variables, and a built-in azure → azure_openai telemetry link routes Azure-via-OpenCode usage onto this tile automatically.
Detection: GROQ_API_KEY environment variable
Tracks rate limits and daily usage windows.
Detection: MISTRAL_API_KEY environment variable
Tracks subscription info and usage endpoints.
Detection: DEEPSEEK_API_KEY environment variable
Tracks rate limits and account balance.
For providers whose billing / usage / account data is gated by web-console session cookies and never exposed via API key, openusage supports a "connect via browser" flow that reads the session cookie directly out of your chosen browser's cookie jar (Chrome / Firefox / Safari / Edge / Brave on macOS / Linux / Windows).
How to connect: Settings → 5 KEYS → navigate to the provider row →
press Enter for browser-session-only providers (for example Perplexity),
or press c on mixed-auth providers (for example OpenCode). Openusage
opens a browser picker, reads the (domain, cookie name) pair declared by
the provider, stores the cookie in credentials.json with 0600
permissions, and uses it on every poll. When the cookie expires, the tile
transitions to AUTH with a "re-login at console.X.com" hint; logging into
the site again in your browser refreshes openusage on the next poll
automatically.
Privacy: opt-in per-account, scoped to a single (domain, cookie name) pair, never sent off-machine. macOS will prompt for Keychain access the first time openusage reads Chrome's cookie store; that's the OS-level consent gate.
Cookie auth currently shipping (full implementation):
- Perplexity →
console.perplexity.ai— tier, balance, spend, analytics - OpenCode →
opencode.ai/_server— balance, monthly limit, subscription
Cookie auth in progress (HAR captured, RPC client needed):
- Google AI Studio →
aistudio.google.com— per-project quotas (needs SAPISIDHASH + MakerSuite tuple decoding; captured 2026-04-30) - ChatGPT consumer →
chatgpt.com— Plus/Team plan + message quotas (HAR captured but thin; needs re-capture from Settings → Subscription pages)
Cookie auth planned (no HAR yet — capture and submit a HAR to enable):
- OpenAI Platform →
platform.openai.com— usage, billing, models - Anthropic Console →
console.anthropic.com— org usage, billing - Mistral Console →
console.mistral.ai— billing, per-model spend - Groq Console →
console.groq.com— usage, billing - xAI Console →
console.x.ai— credit balance, usage breakdown - DeepSeek Platform →
platform.deepseek.com— extended usage history - Z.AI Console →
open.bigmodel.cn— usage detail - Alibaba Cloud Console →
console.aliyun.com— DashScope billing
To add one of these: capture a HAR file from your logged-in browser on
the site (covering the Usage / Billing / Account pages), drop at
~/Downloads/<host>.har, and we wire up the RPC client + parser.
If OpenCode is installed and you've authed any
of its providers, openusage will read ~/.local/share/opencode/auth.json
on startup and adopt the API keys it finds. Currently maps:
| OpenCode entry | openusage account |
|---|---|
moonshotai (api) |
moonshot-ai (provider moonshot) |
openrouter (api) |
openrouter |
zai (api) |
zai |
opencode (api) |
opencode |
ollama-cloud (api) |
ollama-cloud (provider ollama) |
OAuth-typed entries (anthropic, openai, google, cursor) are skipped:
they're chat-scoped tokens, not the API-key shape openusage's poll-time probes
expect. Env-var detection runs first; if both are present the env var wins.
Detection: browser-session cookie from console.perplexity.ai (Settings → 5 KEYS → perplexity → Enter).
Browser-session-auth-only — Perplexity's API key is chat-only. Tile surfaces tier (0–5), available balance, lifetime spend, auto-reload settings, payment method, and 30-day analytics (api_requests, input/output/reasoning tokens, search queries) from the console RPCs at /rest/pplx-api/v2/groups/<org_id>/....
Detection: OPENCODE_API_KEY / ZEN_API_KEY env var for chat-surface auth, optionally a browser-session cookie from opencode.ai for billing data.
Two-tier auth. The API key probes /zen/v1/models for chat-side validation and surfaces the available Zen model count. When connected via browser session (Settings → 5 KEYS → opencode → c), the tile gains balance, monthly limit / monthly usage, auto-reload settings, payment method, and subscription state from the SolidStart server-fn endpoints at opencode.ai/_server.
Openusage auto-discovers the active workspace ID from the authenticated console redirect, so no extra account hint is required for console enrichment.
Detection: MOONSHOT_API_KEY environment variable
Tracks balance breakdown (available_balance = cash_balance + voucher_balance), org-level rate caps (max_request_per_minute, max_token_per_minute, max_concurrency, max_token_quota), tier (user_group_id), and account metadata (org id, project id, masked access key).
By default targets api.moonshot.ai (international, USD). For Moonshot.cn (China, CNY) add a second account in settings.json with "base_url": "https://api.moonshot.cn".
Detection: XAI_API_KEY environment variable
Tracks rate limits and API key info.
Detection: ZAI_API_KEY / ZHIPUAI_API_KEY environment variable, or ~/.chelper/config.yaml
Tracks coding-plan quota limits, model/tool usage, daily trend series, and optional credit balance metadata.
Detection: GEMINI_API_KEY or GOOGLE_API_KEY environment variable
Tracks rate limits and per-model limits.
Detection: ALIBABA_CLOUD_API_KEY environment variable
Tracks quotas, credits, daily usage, and per-model tracking.