TraceKit for AI

AI skills that teach your coding assistant how to set up TraceKit -- live breakpoints, distributed tracing, session replay, and error monitoring.

What is this?

Structured skill files that AI coding assistants read to guide you through setting up TraceKit APM in any project. Each skill contains step-by-step instructions with working code snippets your AI assistant can apply directly.

Supports Claude Code, Cursor, and 38+ other AI tools via npx skills add. Skills follow a detect, configure, verify pattern -- your assistant identifies your stack, sets up TraceKit, and confirms data is flowing to your dashboard.

Compatibility

This repo is the shared source for TraceKit AI integrations. The skills/ content is reused across assistants, but each tool reads a different plugin or marketplace format.

Tool	Skills install	MCP install	Uses shared `skills/`	Metadata source
npx-compatible tools	`npx skills add tracekit-dev/tracekit-for-ai`	Tool-dependent	Yes	repo skills layout
Claude Code	`/install-plugin https://github.com/tracekit-dev/tracekit-for-ai`	Load `.mcp.json` or run `claude mcp add --scope project tracekit ./scripts/run-tracekit-mcp.sh`	Yes	`.claude-plugin/plugin.json` and `.mcp.json`
Cursor	Add `tracekit-dev/tracekit-for-ai` as a plugin source	Load `.cursor/mcp.json`	Yes	`.cursor-plugin/plugin.json` and `.cursor/mcp.json`
Codex	Load `.agents/plugins/marketplace.json` and install `TraceKit`	Run `codex mcp add tracekit -- /absolute/path/to/tracekit-for-ai/scripts/run-tracekit-mcp.sh`	Yes	`.agents/plugins/marketplace.json`, `plugins/tracekit/.codex-plugin/plugin.json`, and `scripts/run-tracekit-mcp.sh`

Short version: one TraceKit repo, one shared skill set, multiple assistant-specific install paths.

Quick Start

npx (works with 38+ AI tools)

npx skills add tracekit-dev/tracekit-for-ai

To install all 23 skills at once (skips the interactive picker):

npx skills add tracekit-dev/tracekit-for-ai --all

This installs the shared skills. MCP availability depends on whether that tool also supports loading this repo's MCP config.

Claude Code

/install-plugin https://github.com/tracekit-dev/tracekit-for-ai

For MCP in Claude Code, either:

claude mcp add --scope project tracekit ./scripts/run-tracekit-mcp.sh

or open this repo's .mcp.json as the project MCP config.

Cursor

Add tracekit-dev/tracekit-for-ai as a plugin source in Cursor settings.

For MCP in Cursor, load .cursor/mcp.json in the project.

Codex

Clone this repo and point Codex at .agents/plugins/marketplace.json, then install the TraceKit plugin from that local marketplace.

For MCP in Codex, add the local launcher script as a global MCP server using an absolute path:

codex mcp add tracekit -- /absolute/path/to/tracekit-for-ai/scripts/run-tracekit-mcp.sh

tracekit is just the server name shown inside Codex. Use an absolute path here because Codex stores the MCP server globally, so relative paths can break when Codex starts from another directory.

On first use, the bundled launcher scripts automatically download the correct tracekit-agent binary from the latest GitHub release into bin/, so end users do not need Go or Python installed.

Authentication

Users should not be sent off to sign up or manually create an API key first if the assistant can walk them through the TraceKit email verification flow.

Auth skill: skills/tracekit-auth/SKILL.md
Auth launcher: scripts/run-tracekit-auth.sh
Unified agent binary entrypoint: scripts/run-tracekit-agent.sh

The helper supports:

./scripts/run-tracekit-auth.sh status
./scripts/run-tracekit-auth.sh register --email <email>
./scripts/run-tracekit-auth.sh verify --session-id <session_id> --code <code>

If the local agent binary is missing, the launcher downloads the correct release binary automatically before running the auth flow.

Successful verification signs the user into the existing account for that email or creates it automatically, then writes the production profile to ~/.tracekitconfig so both the MCP server and future TraceKit skills can reuse the same credentials.

MCP Server

This repo now includes a local TraceKit MCP server for agents that support MCP tool servers.

Shared MCP config: .mcp.json
Cursor MCP config: .cursor/mcp.json
Runner script: scripts/run-tracekit-mcp.sh
Unified agent source: agent/tracekit-agent
Release binaries directory: bin/

The server currently exposes read-focused tools:

tracekit_status
tracekit_dashboard
tracekit_services
tracekit_service_detail
tracekit_traces
tracekit_alert_rules
tracekit_triage_inbox

MCP Auth

The MCP server uses the same TraceKit credentials model as the CLI:

Preferred: the assistant uses tracekit-auth or ./scripts/run-tracekit-auth.sh to write ~/.tracekitconfig
Also supported: tracekit login and stored credentials in ~/.tracekitconfig
Optional override: TRACEKIT_API_KEY, TRACEKIT_USER_ID, and TRACEKIT_ENDPOINT

MCP Requirements

For released binaries in bin/: no language runtime required
If no local binary is present, the launcher self-installs the correct one from the latest GitHub release
For development fallback when no local binary exists: Go 1.22 or newer
valid TraceKit credentials

How To Test

Start the auth flow:

./scripts/run-tracekit-auth.sh register --email you@example.com

Then verify with the emailed code:

./scripts/run-tracekit-auth.sh verify --session-id <session_id> --code <code>

Smoke test the MCP server startup:

./scripts/run-tracekit-mcp.sh

It will wait for MCP stdio messages. Press Ctrl+C to stop it.

Send a manual MCP initialize + tool call:

cd tracekit-for-ai

init='{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2024-11-05","capabilities":{},"clientInfo":{"name":"manual-test","version":"0.0.0"}}}'
call='{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"tracekit_status","arguments":{}}}'

{
  printf 'Content-Length: %s\r\n\r\n%s' ${#init} "$init"
  printf 'Content-Length: %s\r\n\r\n%s' ${#call} "$call"
} | ./scripts/run-tracekit-mcp.sh

Test in Claude Code by loading .mcp.json or running:

claude mcp add --scope project tracekit ./scripts/run-tracekit-mcp.sh

Test in Cursor by loading .cursor/mcp.json for the project.
Test with Codex by:

codex mcp add tracekit -- /absolute/path/to/tracekit-for-ai/scripts/run-tracekit-mcp.sh

Then load .agents/plugins/marketplace.json, install TraceKit, and check that the MCP server is available.

Usage

After installing, ask your AI assistant:

"Set up TraceKit tracing in my Go project"
"Add TraceKit error monitoring to my React app"
"Enable session replay for my Next.js app"

Your assistant reads the relevant skill and walks you through setup.

Available Skills

Backend SDK Setup

Skill	Directory	Description
TraceKit Go SDK	`skills/tracekit-go-sdk/`	Set up distributed tracing and error monitoring in Go services
TraceKit Node.js SDK	`skills/tracekit-node-sdk/`	Set up tracing in Node.js/Express/Fastify/NestJS apps
TraceKit Python SDK	`skills/tracekit-python-sdk/`	Set up tracing in Python/Django/Flask/FastAPI apps
TraceKit PHP SDK	`skills/tracekit-php-sdk/`	Set up tracing in PHP applications
TraceKit Laravel SDK	`skills/tracekit-laravel-sdk/`	Set up tracing in Laravel applications
TraceKit Java SDK	`skills/tracekit-java-sdk/`	Set up tracing in Java/Spring Boot applications
TraceKit .NET SDK	`skills/tracekit-dotnet-sdk/`	Set up tracing in .NET/ASP.NET Core applications
TraceKit Ruby SDK	`skills/tracekit-ruby-sdk/`	Set up tracing in Ruby/Rails applications

Frontend SDK Setup

Skill	Directory	Description
TraceKit Browser SDK	`skills/tracekit-browser-sdk/`	Set up error monitoring and tracing in browser apps
TraceKit React	`skills/tracekit-react-sdk/`	Set up React error boundaries and component tracing
TraceKit Vue	`skills/tracekit-vue-sdk/`	Set up Vue error handler and navigation tracing
TraceKit Angular	`skills/tracekit-angular-sdk/`	Set up Angular ErrorHandler and route tracing
TraceKit Next.js	`skills/tracekit-nextjs-sdk/`	Set up Next.js multi-runtime tracing (server + client + edge)
TraceKit Nuxt	`skills/tracekit-nuxt-sdk/`	Set up Nuxt plugin and server middleware tracing

Feature Setup

Skill	Directory	Description
Code Monitoring	`skills/tracekit-code-monitoring/`	Enable live breakpoints and snapshots
Session Replay	`skills/tracekit-session-replay/`	Record and replay user sessions with privacy controls
Source Maps	`skills/tracekit-source-maps/`	Upload source maps for readable stack traces
Release Tracking	`skills/tracekit-releases/`	Track releases and deploy markers
Alerts	`skills/tracekit-alerts/`	Configure alert rules and notification channels
Distributed Tracing	`skills/tracekit-distributed-tracing/`	Set up frontend-to-backend trace propagation

How Skills Work

Each skill follows a three-step pattern:

Detect -- identifies your language, framework, and package manager from project files
Configure -- provides step-by-step setup with working code snippets your assistant applies directly
Verify -- confirms TraceKit is initialized and sending data to your dashboard

Skills include framework-specific variants, non-negotiable rules (never hardcode API keys, always use env vars), and links to TraceKit docs for advanced configuration.

Links

License

Apache-2.0 -- see LICENSE

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Repository files navigation

TraceKit for AI

What is this?

Compatibility

Quick Start

npx (works with 38+ AI tools)

Claude Code

Cursor

Codex

Authentication

MCP Server

MCP Auth

MCP Requirements

How To Test

Usage

Available Skills

Backend SDK Setup

Frontend SDK Setup

Feature Setup

How Skills Work

Links

License

About

Uh oh!

Releases 7

Packages

Uh oh!

Contributors

Uh oh!

Languages

Name		Name	Last commit message	Last commit date
Latest commit History 21 Commits
.agents/plugins		.agents/plugins
.claude-plugin		.claude-plugin
.cursor-plugin		.cursor-plugin
.cursor		.cursor
.github/workflows		.github/workflows
agent/tracekit-agent		agent/tracekit-agent
bin		bin
plugins/tracekit		plugins/tracekit
scripts		scripts
skills		skills
.gitignore		.gitignore
.mcp.json		.mcp.json
CLAUDE.md		CLAUDE.md
LICENSE		LICENSE
README.md		README.md
SKILL_TREE.md		SKILL_TREE.md

Folders and files

Latest commit

History

Repository files navigation

TraceKit for AI

What is this?

Compatibility

Quick Start

npx (works with 38+ AI tools)

Claude Code

Cursor

Codex

Authentication

MCP Server

MCP Auth

MCP Requirements

How To Test

Usage

Available Skills

Backend SDK Setup

Frontend SDK Setup

Feature Setup

How Skills Work

Links

License

About

Resources

License

Uh oh!

Stars

Watchers

Forks

Releases 7

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

Packages