Traceo bug recordings (console, network, user actions, environment) as context for coding agents.
Scored across 8 tools
Each tool maps to a distinct data slice (comments, console logs, frames, network requests, timeline, user actions), and descriptions explicitly position get_recording as the entry point and get_timeline as the merged view. The only soft overlap is between get_thumbnail and get_frames, both of which return images, though 'single page still' vs 'sequence of stills at moments/ranges' is reasonably distinguishable.
All eight tools use a uniform snake_case get_<resource> pattern with no exceptions or stylistic drift. The naming is predictable enough that an agent can guess a tool's resource from its name alone.
Eight tools is well within the sweet spot and each maps to a genuinely different artifact of a recorded session. No tool feels redundant or like a thin wrapper that could be folded into another.
The read-only investigation surface is well covered: summary, timeline, raw streams, visual frames, and reporter comments for a share link. Minor gaps exist — there is no way to filter/search console logs or requests by predicate (beyond timeline time ranges), and no list/browse of recordings, though the share-link model may make that intentional.
Comments the reporter attached to the recording when sharing it.
Inputs: recording (required).
Browser console output captured during the recording, one line per entry. Secrets are redacted.
Inputs: recording (required), level, limit, cursor, +1 more.
Stills from the recording video, captured after each click, error and failed request plus an even spread. Pass at_s to see the screen at a moment (e.g. the time of an error from get_timeline), or from_s/to_s for a range. Returns up to 6 images.
Inputs: recording (required), at_s, to_s, limit, +1 more.
HTTP requests captured during the recording, one line each: time, method, status, URL (query strings stripped), duration, type, size. No request/response bodies are captured. Sensitive headers are redacted.
Inputs: recording (required), type, limit, cursor, +4 more.
Start here. Summarizes a Traceo bug recording from its share link: page, environment (browser, viewport, locale), duration, event counts, the first errors and failed requests, the user actions leading up to the first error, and which tool to call next.
Inputs: recording (required).
A still image of the recorded page (the recording thumbnail).
Inputs: recording (required).
Console logs, network requests and user actions merged in time order, with times relative to recording start (+12.4s). Best view for reproducing a bug. Narrow with from_s/to_s around an error.
Inputs: recording (required), to_s, limit, types, +2 more.
Clicks, inputs and navigations the reporter performed, in order, with target element and page path.
Inputs: recording (required), limit, cursor.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Production bugs with replay, network, console and stack, ranked for your coding agent to fix.
Bug tracking for AI agents: session replay, console/network logs, repro packs, test plans, evidence.
Jam recordings (video, console, network) as context and proof of work for your coding agent.
Gives your coding agent the captured console, network, replay and screenshot evidence for a bug.
Add one secure layer between your agents and this server.
