ananthunarashimman/tracetel — explained in plain English
Analysis updated 2026-05-18
Send Composio AI-agent tool call activity into Datadog, Grafana Tempo, SigNoz, or Agnost.
Debug what an AI agent's tool calls actually did, including arguments, responses, and timing.
Group related tool calls under one session id to see a full agent run in a dashboard.
| ananthunarashimman/tracetel | 0xradioac7iv/tempfs | 7vignesh/pgpulse | |
|---|---|---|---|
| Stars | 0 | 0 | 0 |
| Language | TypeScript | TypeScript | TypeScript |
| Setup difficulty | easy | moderate | moderate |
| Complexity | 2/5 | 3/5 | 4/5 |
| Audience | developer | developer | developer |
Figures from each repo's GitHub metadata at analysis time.
Requires an existing Composio integration and API key, install as a peer dependency alongside your Composio provider package.
Tracetel is a small TypeScript package that adds observability to Composio, a service that lets AI agents call tools like GitHub, Slack, or Notion on a user's behalf. Normally, when one of those tool calls happens, the only record of it lives inside Composio's own dashboard, and there is no way to pull that data into the monitoring tools a team already uses. Tracetel closes that gap by turning every Composio tool call into a proper OpenTelemetry span, a standard format used across the observability industry, and streaming it to Agnost or any other OpenTelemetry backend such as Datadog, SigNoz, or Grafana Tempo. The package works by wrapping the moment a tool call actually executes. It sits entirely inside Composio's own execution step, not between your app and the AI model, and not between your app and Composio itself. For each tool call, it opens a span before the real API request goes out, records the tool name, the user id, the arguments, and the response, then closes the span once the call finishes. It also captures how long the call took and whether it succeeded or failed. Any argument or response value that looks like a token, password, secret, or API key gets automatically replaced with a redacted placeholder before anything is recorded, so sensitive data never ends up in your traces. There are two ways to use it, matching the two ways Composio itself works: a direct execution mode for one-off tool calls, and a session based mode that also tags every span with a session id, useful for grouping related calls together in a dashboard. Installing it is a matter of running npm install tracetel alongside whichever Composio provider package the project already uses, such as the Vercel AI SDK provider. The main audience is teams already using Composio who want their AI agent's tool activity to show up in the same observability stack they use for the rest of their application, instead of being locked inside a separate dashboard.
Tracetel turns every Composio AI-agent tool call into a standard OpenTelemetry trace, so teams can see what their agent's tools actually did inside their existing monitoring stack.
Mainly TypeScript. The stack also includes TypeScript, Node.js, OpenTelemetry.
License terms are not stated in the README, check the repository's LICENSE file before using in a commercial project.
Setup difficulty is rated easy, with roughly 30min to a first successful run.
Mainly developer.
This repo across BitVibe Labs
double-check against the repo, no cap.