PostHog for OpenClaw
https://github.com/posthog/posthog-openclaw- Category
- Developer Tools
- Rank
- No. 1585Tools index
Previous survey · No. 1590 ·
- Pricing
- Open Source
- Type
- TOOL
- Builder
- PostHog
- GitHub
- 8 stars
- Latest release
- 0.2.5
- Date
About
PostHog LLM analytics plugin for OpenClaw.
What it does
It observes OpenClaw model runs and tool activity, then sends structured generation, span, and trace events to PostHog. It correlates inputs with outputs, records tokens, latency, costs, errors, models, providers, sessions, channels, and agents.
Why it's ranked here
This is a focused observability bridge with unusually clear event coverage. Message and session trace grouping, tool parenting, content redaction, and session windows make the captured data useful beyond raw logs. Its value depends entirely on running OpenClaw and using PostHog.
What's good
Privacy mode removes prompts, responses, and tool parameters while preserving operational metrics. Input normalization converts several OpenClaw and Anthropic message shapes into PostHog-compatible chat records, strips reasoning blocks, and preserves tool-call relationships. Tests cover configuration precedence, redaction, trace reuse, event fields, and lifecycle behavior.
Tradeoffs
Trace-level events require OpenClaw diagnostics to be enabled. Privacy mode still sends token counts, latency, model information, and error status. The plugin needs a PostHog project key, Node 20 or newer, and a compatible OpenClaw installation. Its OpenClaw type definitions are copied from upstream pending stable exports.
How to use it well
Use it when an OpenClaw deployment needs generation costs, latency, failures, tool timing, and conversation traces inside PostHog. Choose message grouping for per-request analysis or session grouping for longer journeys. Enable privacy mode for content-sensitive workloads. It does not replace logging, alerting, or general infrastructure monitoring.
Technical notes+
package.json defines an ESM npm package named @posthog/openclaw, requires Node >=20.0.0, depends on posthog-node, and lists OpenClaw as an optional peer. index.ts resolves configuration and environment fallbacks before registering hooks. src/plugin.ts lazily creates the PostHog client, subscribes to message.processed diagnostics, tracks runs and session windows in Maps, captures lifecycle hooks, and flushes on service shutdown or process exit. src/events.ts builds $ai_generation, $ai_span, and $ai_trace payloads. src/utils.ts normalizes message formats, removes thinking and reasoning blocks, and applies content redaction. Tests are co-located under src, while scripts/verify-package.mjs checks the packed artifact and excludes test files.
Observed
- License
- MIT
- Primary language
- TypeScript
- Packaging
- ES module published to npm as @posthog/openclaw
- Install surface
- Installed through the OpenClaw plugin installer
- Interface
- OpenClaw plugin using model input, model output, tool completion, and diagnostic event hooks
- Runtime support
- Node.js 20 or newer
- Dependency surface
- Uses posthog-node and declares OpenClaw as an optional peer dependency
- Test structure
- Vitest tests are co-located with source files under src
Read from README.md, package.json, index.ts, src/types.ts, src/utils.ts, src/events.ts, src/plugin.ts, src/version.ts, src/index.test.ts, src/events.test.ts, src/plugin.test.ts, src/openclaw-types.ts, src/openclaw-plugin-sdk.d.ts, vitest.config.ts, scripts/verify-package.mjs.
What it can do
Track LLM API calls and usage
LLM API requests and responses → Analytics data on API usage patterns
Monitor LLM model performance metrics
LLM model execution data → Performance metrics and statistics
Analyze LLM token consumption
LLM token usage data → Token consumption reports and trends
Generate LLM cost analytics
LLM usage and pricing data → Cost breakdowns and billing insights
Track user interactions with LLM features
User session and interaction data → User behavior analytics for LLM usage
Tags
Tech Stack
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.