# Tapestry Tapestry is a user/agent support and reinforcement system. These docs describe how the platform observes coordination between operators and agents and how durable structure is produced from that observation. ## Overview - [Tapestry](https://tapestry-khaki.vercel.app/docs/): A user/agent support and reinforcement system. The mechanisms — memory, telemetry, observability, architecture analysis, friction analysis, upskilling — observe, strengthen, and stabilize coordination between the operator and agents. ## Start - [First Observatory visit](https://tapestry-khaki.vercel.app/start/first-observatory-visit/): Open the dashboard, learn what's normal for a brand-new project (empty inbox, sparse signals), and hand off to the Observatory section for deeper reading. - [Verify it worked](https://tapestry-khaki.vercel.app/start/verify-it-worked/): Four quick checks confirming the platform is wired correctly. Each check has a one-line fix path when it fails, plus a link into the relevant Systems page for deeper troubleshooting. - [What Tapestry is](https://tapestry-khaki.vercel.app/start/what-stays-on-track/): An intelligence that watches across all your projects and turns recurring friction into durable structure — so your work keeps improving even though the agent itself only changes when a new model ships. - [What Tapestry is not](https://tapestry-khaki.vercel.app/start/what-tapestry-is-not/): Tapestry borrows pieces from memory tools, observability stacks, and tracing — but it is none of them. Here's what it actually is, frame by frame. - [Your first project](https://tapestry-khaki.vercel.app/start/your-first-project/): What `tapestry onboard` actually did, what files appeared in your repo, and what to expect when you start your first Claude Code session. Run this page after the Quickstart and before Verify it worked. ## How To - [Connect VS Code (without Claude Code)](https://tapestry-khaki.vercel.app/how-to/connect-vscode-mcp/): Add Tapestry's memory and docs MCP servers to VS Code's native MCP support, so Copilot Chat and other MCP-aware assistants can use them — no Claude Code, no extension required. - [Create your own plugin](https://tapestry-khaki.vercel.app/how-to/create-your-own-plugin/): Tapestry ships one universal discipline plugin. When a project's shape needs more — project-specific guards, or your own memory/observer/telemetry backend — generate your own personalized plugin pointed at your own deployment and publish it to your own marketplace. - [Quickstart — VS Code](https://tapestry-khaki.vercel.app/how-to/quickstart-vscode/): 4-step setup for a new Tapestry-consuming project in VS Code. Uses the Claude Code extension + the `tapestry onboard` CLI to bootstrap configuration in one command. - [Recover from common failures](https://tapestry-khaki.vercel.app/how-to/recover-from-common-failures/): Symptom-to-fix table for the failure modes that happen when a discipline-stack piece goes missing. - [Set up a new project](https://tapestry-khaki.vercel.app/how-to/set-up-a-new-project/): Step-by-step checklist to wire a new project so the agent in it gets the full discipline stack from day one. - [Set up Claude Code](https://tapestry-khaki.vercel.app/how-to/set-up-claude-code/): Install Claude Code, sign in, and connect the Tapestry marketplace so the discipline and patterns plugins are available in your sessions. - [Set up GitHub](https://tapestry-khaki.vercel.app/how-to/set-up-github/): A GitHub account and access — for the Tapestry plugin marketplace, for connecting Render and Vercel to your deployment repo, and for publishing your own plugins. - [Set up Grafana Cloud + OTel](https://tapestry-khaki.vercel.app/how-to/set-up-grafana-cloud/): Step-by-step walkthrough for the telemetry pipeline — sign up for Grafana Cloud, get OTLP credentials, and wire OTEL_EXPORTER_OTLP_* env vars so the discipline plugin's hooks emit to Grafana Loki/Tempo. Required for cross-machine observability. - [Set up Render](https://tapestry-khaki.vercel.app/how-to/set-up-render/): Step-by-step walkthrough for provisioning Tapestry's backend services on Render — managed Postgres + the FastAPI services declared in the platform's `render.yaml` Blueprint. You run this once for your own deployment. - [Set up Vercel](https://tapestry-khaki.vercel.app/how-to/set-up-vercel/): Step-by-step walkthrough for deploying Tapestry's frontend on Vercel — the Astro Starlight docs site (this site) and the Observatory cockpit. You run this once per surface for your own deployment. ## Explanation - [Architecture snapshots](https://tapestry-khaki.vercel.app/explanation/architecture-snapshots/): What the architecture-snapshot automation is, what it captures, when it fires, and how to recover when it goes missing. - [The discipline stack](https://tapestry-khaki.vercel.app/explanation/discipline-stack/): Why Tapestry's mechanisms exist — each is a structural reinforcement against a specific kind of weak bond in the user-agent interface. Orientation map of the discipline stack with links to deep-dive pages. - [The memory MCP and your project's memory](https://tapestry-khaki.vercel.app/explanation/memory-mcp/): What the loom-memory MCP is, what gets stored in it, how the memory for your project stays scoped and organized, and what you can do as the operator to keep it healthy. - [The plugins](https://tapestry-khaki.vercel.app/explanation/plugins/): Every Claude Code plugin that's part of the Tapestry platform — what each does, how they interact, the install/enable/restart lifecycle, and how to keep them clean and updated. - [The observer](https://tapestry-khaki.vercel.app/explanation/the-observer/): How Tapestry notices what keeps recurring in your project — the correction you make three sessions running, the workflow you've now done the same way five times — and turns the recurring things into durable structure instead of leaving you to re-explain them. ## Systems - [Docs MCP](https://tapestry-khaki.vercel.app/systems/docs-mcp/): Three ways to grab the Tapestry docs as machine-readable content — a static llms.txt corpus, per-page raw markdown, and a local stdio MCP server. No hosted service required. - [Memory](https://tapestry-khaki.vercel.app/systems/memory/): The MCP server that holds typed memory records and exposes them as memory_read / memory_write / memory_recall / memory_search / memory_list / memory_delete tools to any agent. The substrate every other component reads and writes. - [Observatory](https://tapestry-khaki.vercel.app/systems/observatory/): The component view of the Observatory — how it's wired to Memory and the registries, how to self-host it, verify it, and troubleshoot it. For what it is and how to read it, see the Learn page. - [Observer](https://tapestry-khaki.vercel.app/systems/observer/): The component that interprets signals into patterns. Runs as a scheduled cron and as an on-demand subagent. The interpretation layer of the Signal → Interpretation → Pattern pipeline. - [Registry](https://tapestry-khaki.vercel.app/systems/registry/): Two services that hold the platform's pattern store. The Architecture Registry stores durable structural facts. The Candidate Registry stores patterns the Observer has surfaced but not yet promoted. Together they're the pre-promotion + post-promotion sides of the same governance flow. - [Telemetry](https://tapestry-khaki.vercel.app/systems/telemetry/): The OTel pipeline that carries signals from hooks and runtime services to Grafana Cloud. Producer side is the discipline plugin's hooks; transport is OTLP/HTTP; destination is Grafana Cloud (Loki for logs, Tempo for traces). Without it, no cross-machine observability. ## Observatory - [The Observatory](https://tapestry-khaki.vercel.app/observatory/about/): The console where you watch coordination between you and your agents over time — what's changing in a project, what the observer sees, where friction appears, and whether things are trending better or worse. It lives at /observatory. - [The Observatory feed](https://tapestry-khaki.vercel.app/observatory/feed/): The data contract behind the console — the /api/episodes.json endpoint, the sources it reads from, the event shape a live source must provide, and how to point the console at a central coordination store. - [Run the Observatory](https://tapestry-khaki.vercel.app/observatory/run-it/): Open the Observatory console with the tapestry CLI, or run it locally for development. Includes the deployment settings the console needs. ## Reference - [Load-bearing files](https://tapestry-khaki.vercel.app/reference/load-bearing-files/): Every file, config block, and environment variable the discipline stack depends on, with a one-line "why it exists" and what fails if it's missing. - [OTel coordination contract](https://tapestry-khaki.vercel.app/reference/otel-coordination-contract/): The attribute schema that Tapestry-emitted OTel events must carry so the observer can interpret them as coordination signals. OTel is the transport. Tapestry interprets the signal as coordination reinforcement. - [Platform dependencies](https://tapestry-khaki.vercel.app/reference/platform-dependencies/): The external services Tapestry runs on. What's required to operate the platform vs what's required to use it from a consuming project.