# Wolbarg Workspace — Full Documentation > Complete Markdown export of Workspace docs for AI systems. > Prefer /llms.txt for a curated index. Site: https://workspace.wolbarg.com Generated from 7 documentation pages. Primary audience: Claude Code users connecting `@wolbarg/workspace` via MCP. --- # FAQ > Common questions about Wolbarg Workspace. URL: /docs/faq ## What is Wolbarg Workspace? [#what-is-wolbarg-workspace] [`@wolbarg/workspace`](https://www.npmjs.com/package/@wolbarg/workspace) is the product layer for coding agents (Cursor, Claude Code, Codex). It keeps shared, reconciled project truth on top of the `wolbarg` SDK. ## Is this the same as the Wolbarg SDK? [#is-this-the-same-as-the-wolbarg-sdk] No. The [SDK](https://wolbarg.com) (`wolbarg`) is for builders wiring memory into custom agents. Workspace is the ready-made product for coding-agent hosts. ## Do I need to edit MCP JSON by hand? [#do-i-need-to-edit-mcp-json-by-hand] No. `npx @wolbarg/workspace connect ` installs MCP + host rules for you. ## Where is data stored? [#where-is-data-stored] Locally by default (SQLite under `.wolbarg/workspace/`). Postgres is optional via the `init` wizard. ## Can Cursor and Claude Code share memory? [#can-cursor-and-claude-code-share-memory] Yes — that is the point. Connect both hosts to the same project workspace. --- # Wolbarg Workspace > Shared project truth for Claude Code — also Cursor and Codex. Install with @wolbarg/workspace. URL: /docs Built for **Claude Code** first. Also connects Cursor and Codex. The product sits on top of the [`wolbarg`](https://wolbarg.com/docs/quick-start) SDK. ## Why use it? [#why-use-it] Claude Code forgets when a session ends. It re-investigates the same code, reopens settled decisions, and acts on stale context. Wolbarg Workspace keeps one shared, always-current view of your project — decisions, architecture, tasks, bugs, conventions — and every connected agent reads and updates it automatically. ## Quick start (Claude Code) [#quick-start-claude-code] ```bash npm install @wolbarg/workspace npx @wolbarg/workspace init npx @wolbarg/workspace connect claude ``` Then restart Claude Code. On the first session, paste the [seed prompt](/docs/install#seed-project-memory-first-session) so the agent walks the repo and stores current decisions into Workspace. See [Install](/docs/install) and [Claude Code](/docs/connectors/claude-code). ## Also supported [#also-supported] | Host | Command | Docs | | ----------- | ---------------- | ------------------------------------- | | Claude Code | `connect claude` | [Setup](/docs/connectors/claude-code) | | Cursor | `connect cursor` | [Setup](/docs/connectors/cursor) | | Codex | `connect codex` | [Setup](/docs/connectors/codex) | ## Packages [#packages] | Package | Purpose | | -------------------------------------------------------------------------------- | ----------------------------------- | | [`@wolbarg/workspace`](https://www.npmjs.com/package/@wolbarg/workspace) | CLI + domain library | | [`@wolbarg/workspace-mcp`](https://www.npmjs.com/package/@wolbarg/workspace-mcp) | MCP server (installed by `connect`) | Source: [github.com/wolbarg/workspace](https://github.com/wolbarg/workspace) --- # Install > Install and initialize Wolbarg Workspace in your project with @wolbarg/workspace. URL: /docs/install ## Install [#install] Published on npm as **`@wolbarg/workspace`**. Run inside your project folder (not global): ```bash npm install @wolbarg/workspace npx @wolbarg/workspace init ``` A short wizard asks for storage and an embedding provider. Press Enter to keep the defaults (SQLite + Ollama, fully local). Skip the wizard: ```bash npx @wolbarg/workspace init --yes ``` ## Embedding provider [#embedding-provider] Workspace needs an embedding model to search memory. The default is [Ollama](https://ollama.com): ```bash ollama pull nomic-embed-text ``` The `init` wizard also offers OpenAI, Voyage AI, Gemini, LM Studio, OpenRouter, or any OpenAI-compatible endpoint. API keys go in `.env.wolbarg` — never commit that file. ## Config [#config] `init` writes: | File | What it is | | -------------------------------- | ------------------------------- | | `.wolbarg/workspace/config.json` | Storage + embedding settings | | `.wolbarg/workspace/memory.db` | Local SQLite database (default) | | `.env.wolbarg` | API keys — do not commit | Defaults: SQLite storage, Ollama embeddings, 90-day retention. Teams can switch to Postgres in the wizard. Re-run `init` any time to change providers. ## Check that it works [#check-that-it-works] ```bash npx @wolbarg/workspace doctor npx @wolbarg/workspace status npx @wolbarg/workspace review ``` ## Connect an agent [#connect-an-agent] ```bash npx @wolbarg/workspace connect claude npx @wolbarg/workspace connect cursor npx @wolbarg/workspace connect codex ``` Restart the agent afterwards. See [Connectors](/connectors) for host-specific details. ## Seed project memory (first session) [#seed-project-memory-first-session] After `init` + `connect`, paste this into **Claude Code**, **Cursor**, or **Codex**. The agent should walk the repo and write present truth into Workspace (SQLite or Postgres — whichever you chose in `init`) via `memory_write`. ```text You are seeding Wolbarg Workspace for this project. Goal: explore the codebase and store the current project truth into Wolbarg Workspace memory (the configured SQLite or Postgres store). Do not invent facts. Prefer evidence from files, tests, and docs. Steps: 1. Confirm Wolbarg MCP tools are available. Call `memory_load` first. 2. Map the project: package manifests, README, docs/, architecture notes, main entrypoints, app/package layout, CI configs, env examples (names only — never store secret values). 3. Extract present truth only — not a chat transcript. Cover: - architecture.* — how the system is structured, key modules, data flow - decision.* — settled product/tech choices and why - constraint.* — hard rules (Node version, no force-push, local-first, etc.) - convention.* — naming, folder layout, commit/PR norms - glossary.* — project-specific terms - task.* — known open work (if clearly documented) - bug.* — known open bugs (if clearly documented) - finding.* — important discoveries that future agents should not re-investigate 4. For each fact, call `memory_write` with: - kind: one of architecture | decision | constraint | convention | glossary | task | bug | finding - key: stable namespaced id, e.g. decision.auth.session-strategy - title + statement (statement ≤ 240 chars, present tense) - detail when useful - evidence: file/doc/test/command refs with paths - scope.paths / scope.components when relevant - confidence 0–1 based on how certain the evidence is 5. Batch related observations (up to 50 per `memory_write`). Prefer fewer high-signal facts over dumping every file. 6. Never store secrets, API keys, tokens, .env values, passwords, or raw logs. 7. When done, call `memory_reconcile` with mode "deep", then `memory_load` and summarize what was stored. Start now. Explore the repo thoroughly, then write memory. ``` After it finishes, check: ```bash npx @wolbarg/workspace status npx @wolbarg/workspace review ``` --- # Claude Code > Connect Wolbarg Workspace to Claude Code — the primary host path for shared project memory. URL: /docs/connectors/claude-code ## Connect [#connect] From your project root (after [`init`](/docs/install)): ```bash npx @wolbarg/workspace connect claude ``` Restart Claude Code afterwards. Decisions and discoveries persist across sessions and stay available to Cursor / Codex if those hosts are connected too. ## What you get [#what-you-get] * Shared project memory tools (`memory_load`, `memory_search`, `memory_write`, …) * Project-scoped truth that survives Claude Code session ends * No hand-edited MCP JSON ## Verify [#verify] ```bash npx @wolbarg/workspace doctor npx @wolbarg/workspace status ``` ## Seed memory (recommended) [#seed-memory-recommended] On the first connected session, paste the **Seed project memory** prompt from [Install](/docs/install#seed-project-memory-first-session). Claude Code will scan the repo and write architecture, decisions, constraints, and conventions into Workspace. ## Related [#related] * [Install](/docs/install) * [Cursor](/docs/connectors/cursor) * [Codex](/docs/connectors/codex) * [All connectors](/connectors) --- # Codex > Connect Wolbarg Workspace to Codex CLI for durable project context. URL: /docs/connectors/codex ## Connect [#connect] From your project root (after [`init`](/docs/install)): ```bash npx @wolbarg/workspace connect codex ``` Restart Codex afterwards. The same reconciled project truth is shared with Cursor and Claude Code. ## What you get [#what-you-get] * Durable project memory across Codex sessions * Shared decisions, tasks, bugs, and conventions with other hosts * One CLI — no manual MCP config editing ## Verify [#verify] ```bash npx @wolbarg/workspace doctor npx @wolbarg/workspace status ``` ## Seed memory (recommended) [#seed-memory-recommended] On the first connected session, paste the **Seed project memory** prompt from [Install](/docs/install#seed-project-memory-first-session). Codex will scan the repo and write current project truth into Workspace. ## Related [#related] * [Install](/docs/install) * [Cursor](/docs/connectors/cursor) * [Claude Code](/docs/connectors/claude-code) * [All connectors](/connectors) --- # Cursor > Connect Wolbarg Workspace to Cursor with MCP tools and host rules. URL: /docs/connectors/cursor ## Connect [#connect] From your project root (after [`init`](/docs/install)): ```bash npx @wolbarg/workspace connect cursor ``` This installs the Workspace MCP server and Cursor host rules for the current project. Restart Cursor afterwards. ## What you get [#what-you-get] * Shared project memory tools (`memory_load`, `memory_search`, `memory_write`, …) * Cursor rules that tell the agent to use Workspace as current project truth * No hand-edited MCP JSON ## Verify [#verify] ```bash npx @wolbarg/workspace doctor npx @wolbarg/workspace status ``` ## Seed memory (recommended) [#seed-memory-recommended] On the first connected session, paste the **Seed project memory** prompt from [Install](/docs/install#seed-project-memory-first-session). Cursor will scan the repo and write current project truth into Workspace. ## Related [#related] * [Install](/docs/install) * [Claude Code](/docs/connectors/claude-code) * [Codex](/docs/connectors/codex) * [All connectors](/connectors) --- # Connectors > Connect Wolbarg Workspace to Cursor, Claude Code, and Codex. URL: /docs/connectors ## Product page [#product-page] Browse every coding-agent connector: **[→ All connectors](/connectors)** ## Hosts [#hosts] | Host | Command | Docs | | ----------- | --------------------------------------- | ------------------------------------- | | Cursor | `npx @wolbarg/workspace connect cursor` | [Setup](/docs/connectors/cursor) | | Claude Code | `npx @wolbarg/workspace connect claude` | [Setup](/docs/connectors/claude-code) | | Codex | `npx @wolbarg/workspace connect codex` | [Setup](/docs/connectors/codex) | Restart the agent after connecting. No manual MCP JSON editing is required. ## Building custom agents? [#building-custom-agents] For AI SDK adapters (Vercel AI, LangChain, OpenAI Agents, and more), see the [Wolbarg SDK site](https://wolbarg.com/docs/integrations).