Setup by the agent, not the docs
/traceway-setup picks the right path per part of the stack — OTel for backends, SDKs for browser and mobile — and covers background tasks, AI traces, and source maps.
Specialized · Agent Skills
AI-first observability. Your agent does the debugging.
Your agent sets up Traceway, queries production telemetry, and finds the root cause. Works with Claude Code, Cursor, Codex, and any agent that reads SKILL.md.
Service facts
What one install hands your agent
Two playbooks and an agent-first CLI.
- 2skills, one install
- /traceway-setup wires the stack; /traceway queries and debugs it. Plain Markdown, no marketplace.
- 1mutating action
- Archiving an exception group — and it requires an explicit --yes. Everything else is read-only.
- 6+agents supported
- Claude Code, Cursor, Codex, OpenCode, Gemini CLI, Copilot — anything that reads SKILL.md.
Open format
Plain Markdown. No marketplace. No lock-in.
The skills live in the same MIT-licensed repo as Traceway itself, and Cloud is convenience, not a gated platform — self-host and the skills work against your instance with traceway login --url.
npx skills add tracewayapp/tracewayInstalls the skills for every SKILL.md-reading agent it finds on your machine.
- /traceway-setup
- Wires the stackDetects frameworks from the repo; OTel for backends, Traceway SDKs for web and mobile; verifies data arrives grouped.
- /traceway
- Debugs itInstalls and authenticates the CLI, then queries exceptions, logs, endpoints, and metrics to root cause.
- CLI output
- JSON when pipedTables on a TTY, --fields trims responses, stable error identifiers and exit codes.
- Mutations
- Require --yesNothing hangs waiting for input; the CLI is read-only apart from archiving.
On this line
From bug report to root cause.
Paste the report as written. The agent pulls the matching exceptions, the logs around the failure, and endpoint stats, then opens the failing code.
Telemetry the agent reads itself
Grouped exceptions with full stack traces, logs correlated by trace ID, and first-seen times that point at the breaking deploy.
An agent-first CLI
Built for non-interactive use: JSON output, stable errors, nothing that hangs. Humans on a TTY still get tables.
Ends with a proposed fix
The debug skill walks the trace to the failing code and leaves a patch ready for review in your repo.
Safe by default
The agent can't hurt production.
The CLI is read-only apart from archiving exception groups, and that requires an explicit --yes flag. Works against self-hosted instances too — profiles let one machine use several.
Next stop
Give your agent production context.
Install the skills and let your agent take the next bug.