Set up with AI

If you use an AI coding agent such as Claude Code, Codex or Cursor, you can hand it the entire Errorgap setup. Errorgap generates a single prompt that contains your project credentials and points the agent at step-by-step install docs for your stack. The agent installs the SDK, wires the configuration, and proves the connection works by sending a test event.

How it works

1. Create a project in Errorgap and open its setup screen.
2. In the configure step, switch from Manual to Set up with AI.
3. Copy the generated prompt and paste it into your agent, run from your application's repository.
4. The setup screen advances automatically when the first event arrives.

Copy the prompt

The prompt is generated per project and already includes the project API key, the endpoint of your Errorgap instance, and the project slug. There is nothing to fill in. It instructs the agent to fetch the install doc for the stack you selected, so pick your stack before switching to the AI option.

What the agent does

• Fetches the machine-readable install doc for your stack from your Errorgap instance.
• Installs the SDK package and adds the configuration file, keeping the API key in an environment variable.
• Verifies connectivity by posting a test event directly to the ingestion API and checking for an HTTP 201 response.
• Triggers a real error through the installed SDK to confirm the in-app integration works.
• Asks which other Errorgap features you want, such as deploy tracking, source maps, performance, logs, session replay and uptime, then sets up the ones you choose.

The setup skill

The questions come from the Errorgap setup skill, a pair of markdown files that the prompt tells the agent to fetch: /skills/errorgap-setup/SKILL.md and /skills/errorgap-setup/features.md. The skill has the agent inspect your repository first, so it only asks about features that apply, and it recommends an answer for each question. Steps that need you, such as creating an uptime monitor or connecting Slack, come back as one checklist at the end.

To add a feature later without a new prompt, install the skill in your repository. Claude Code reads skills from .claude/skills and Codex from .agents/skills:

for dir in .claude/skills .agents/skills; do
  mkdir -p "$dir/errorgap-setup"
  for f in SKILL.md features.md; do
    curl -fsSL "https://errorgap.com/skills/errorgap-setup/$f" -o "$dir/errorgap-setup/$f"
  done
done

Then run /errorgap-setup in Claude Code or $errorgap-setup in Codex, or ask the agent to set up an Errorgap feature. The skill holds no credentials. It reads them from your environment or asks for them.

Machine-readable install docs

Each Errorgap instance serves agent-oriented markdown guides at /agent-docs/index.md, with one guide per supported stack (Rails, Rack, Node.js, Django, Flask, Laravel, WordPress, Go, and plain HTTP). The index documents the ingestion API, the auth header, and the verification request:

curl -sS -X POST "$ERRORGAP_ENDPOINT/api/projects/$ERRORGAP_PROJECT_SLUG/notices" \
  -H "content-type: application/json" \
  -H "x-errorgap-project-key: $ERRORGAP_API_KEY" \
  -d '{"errors":[{"type":"ErrorgapInstallTest","message":"Errorgap install verification"}]}'

Because the docs are served by Errorgap, the generated prompt and install instructions stay aligned with the project and stack you selected.

Security notes

• The prompt contains a live project API key. Treat it like a secret and avoid pasting it into shared or logged channels.
• The generated instructions tell the agent to keep the key in an environment variable and never commit it.
• You can rotate the project key at any time in project settings if it is exposed.