Skip to content
Tapestry Tapestry

Your first project

Copy page

You ran tapestry onboard my-project. Here’s what landed and what comes next.

Files written

Section titled “Files written”

tapestry onboard is idempotent and writes (or merges into) four files:

FileWhat it holdsWhy
.envLOOM_PROJECT_ID=my-project + shared OTel + API-key blocksThe project’s identity tag — everything memory and telemetry scope by this.
.mcp.jsonThe loom-memory MCP server declarationWires the Memory MCP into every Claude Code session in this repo.
.project-intelligence/my-project/agent-profile.json, project-context.json, observatory-config.jsonPer-project intelligence files the Observer + Observatory read.
.claude/settings.jsonBoth tapestry-discipline + tapestry-patterns plugins enabledSo the discipline plugin auto-loads its hooks and tapestry-patterns auto-loads its agents + skills.

It also registers your project with the Project Registry if the registry is reachable. If it isn’t, the local files still write — registration retries on the next tapestry onboard run.

What changes in Claude Code

Section titled “What changes in Claude Code”

Open the chat panel and start a fresh session. Three things are now different from a bare project:

1. The session header includes an auto-recall block. It lists memory records tagged for my-project. On first session this block is empty — that’s expected; it fills in as you accumulate corrections, decisions, and observer findings over time.

2. Every prompt you send gets a [loom-discipline] Discipline check: reminder line injected as additional context. That’s the discipline plugin’s UserPromptSubmit hook firing — it injects PROBE-before-asserting + cite file:line + dev-tooling-vs-runtime + save-friction-as-memory rules into the agent’s working context.

3. Hook events emit to two destinations. Local ~/.claude/logs/hooks.jsonl always (source of truth). Grafana Cloud via OTel if OTEL_EXPORTER_OTLP_* env vars are set in your .env. See Telemetry for the wiring.

What’s NOT installed by onboarding

Section titled “What’s NOT installed by onboarding”

Onboarding makes your project observable. It does not install a per-project Observatory — the Observatory is one surface in your platform deployment, shared across all your projects. Onboarding wires your project into it.

What to do next

Section titled “What to do next”
  1. Verify it worked — four quick checks that the wiring landed.
  2. First Observatory visit — open the dashboard and learn what to look for.
  3. (Optional) Write a CLAUDE.md in your project root. Onboarding does not author one — it’s project-specific. See the example in the project’s CLAUDE.md.
Section titled “Related”