Package Exports

opencode-telemetry
opencode-telemetry/src/index.ts

This package does not declare an exports field, so the exports above have been automatically detected and optimized by JSPM instead. If any package subpath is missing, it is recommended to post an issue to the original package (opencode-telemetry) to support the "exports" field. If that is not possible, create a JSPM override to customize the exports field for this package.

Readme

opencode-telemetry

Passively logs per-turn telemetry for opencode sessions into a local SQLite database. Zero configuration required.

What it does

Records every assistant turn: tokens (input/output/cached/reasoning), model, provider, agent, latency, finish reason, and estimated cost.
Records every tool call: name, skill name (for skill tool), args size, result size, duration, and status.
Tracks session metadata: project path, parent session (for subagents), start/end time, and aggregate totals.
Provides /telemetry-report and /telemetry-inspect slash commands for quick analysis from within opencode.

What it does NOT do

No message body logging. Prompt content and tool result text are never stored — only byte sizes and token counts.
No network calls. All data stays on your machine. Zero outbound requests, ever.
No cloud, no accounts, no real-time UI. The SQLite file is the API.

Install

npm install opencode-telemetry

Add to your opencode.json:

{
  "plugin": [
    "opencode-telemetry"
  ]
}

Restart opencode. The database is created automatically on the first session event — no setup needed.

Usage

Slash commands (inside opencode)

/telemetry-report

Outputs a markdown report covering the last 7 days: headline stats, top sessions by cost, per-agent and per-model breakdowns, skill usage, largest tool results, and cache efficiency.

/telemetry-inspect <session_id>

Deep-dives into a single session: turn-by-turn metrics, tool call timeline, skill load summary, and cost breakdown.

Direct SQL access

The database lives at ~/.local/share/opencode-telemetry/data.db (Linux/macOS) or %LOCALAPPDATA%\opencode-telemetry\data.db (Windows).

Pre-canned queries are in the queries/ directory:

DB=~/.local/share/opencode-telemetry/data.db

# Top sessions by cost this week
sqlite3 $DB < queries/top-consumers-7d.sql

# Context bloat suspects (high input/output ratio)
sqlite3 $DB < queries/ratio-in-out-by-agent.sql

# Skills loaded multiple times in the same session
sqlite3 $DB < queries/duplicate-skills.sql

# Largest tool result payloads
sqlite3 $DB < queries/largest-tool-results.sql

# Cache hit rate by model
sqlite3 $DB < queries/cache-efficiency.sql

# Daily token trend (last 30 days)
sqlite3 $DB < queries/daily-token-trend.sql

Or any ad-hoc SQL:

sqlite3 ~/.local/share/opencode-telemetry/data.db \
  "SELECT model, SUM(input_tokens+output_tokens) AS tok FROM turns GROUP BY model ORDER BY tok DESC;"

Schema

Three main tables — see src/db.ts for the full DDL.

Table	One row per…
`sessions`	opencode session
`turns`	assistant message (= one API call)
`tool_calls`	tool invocation

Plus _meta for schema versioning.

Pricing accuracy

src/pricing.json is a static snapshot (last updated April 2026). It covers Anthropic Claude (Opus/Sonnet/Haiku), OpenAI (GPT-4.5, GPT-4.1, o3, o4-mini), Google Gemini (2.5 Pro/Flash), and common local models (at $0). If a model is not in the table, est_cost_usd is stored as NULL — never fabricated.

Rates drift over time. PRs to update pricing.json are welcome.

Privacy

Local only. The plugin makes no outbound network calls.
No prompt content stored. Only byte sizes, token counts, timings, and structural metadata are persisted.
No telemetry phone-home. The plugin itself is not instrumented.

Roadmap

Auto-cleanup TTL (delete sessions older than N days)
More pre-canned queries (anomaly detection, cost forecasting)
Web dashboard (if there's demand)

Contributing

PRs welcome — especially for pricing.json updates and new queries. Please keep the plugin itself under ~800 lines of source.

License

MIT