# WebAccess.MCP WebAccess.MCP is a Streamable HTTP MCP server that provides web search and Playwright-backed browser tools to LM Studio. ## Start After Copying or Cloning the Repository These instructions assume the computer already has: - Podman Desktop. - A running Podman machine. - Podman Compose. - LM Studio 0.3.17 or later. - A tool-capable model installed in LM Studio. Confirm that `Containerfile` exists directly in the copied repository root. Open **PowerShell (Standard user)** in the repository root. Do not run PowerShell as administrator. Run the PowerShell startup script: ```powershell powershell.exe -NoProfile -ExecutionPolicy Bypass -File ".\deploy\Start-WebAccessStack.ps1" ``` Alternatively, run the shorter wrapper: ```powershell .\Start-WebAccessMcp.cmd ``` Both commands: 1. Verify that Podman is installed and its machine is running. 2. Verify that Podman Compose is installed. 3. Create `deploy\.env` with machine-specific secrets if it is missing. 4. Build the WebAccess.MCP container image. 5. Start PostgreSQL, SearXNG, and WebAccess.MCP. 6. Wait for the WebAccess.MCP health endpoint. 7. Print container status, endpoint URLs, and the LM Studio configuration. The first build can take several minutes. ## Verify WebAccess.MCP Use **PowerShell (Standard user)**: ```powershell curl.exe http://127.0.0.1:28081/health ``` Expected response: ```json {"status":"healthy"} ``` The MCP endpoint is: ```text http://127.0.0.1:28081/mcp ``` A plain GET request to `/mcp` can return HTTP 405 because it is an MCP protocol endpoint, not a normal web page. Use `/health` for connectivity checks and LM Studio for the MCP handshake. ## Configure LM Studio No `mcp.json` change is needed if it already contains this exact server entry: ```json { "mcpServers": { "webaccess": { "url": "http://127.0.0.1:28081/mcp" } } } ``` If `mcp.json` already contains other MCP servers: 1. Open LM Studio. 2. Open the **Program** tab. 3. Select **Install > Edit mcp.json**. 4. Keep the existing server entries. 5. Add this property inside the existing `mcpServers` object: ```json "webaccess": { "url": "http://127.0.0.1:28081/mcp" } ``` 6. Include a comma between server properties. 7. Save `mcp.json`. 8. Click **Refresh** in LM Studio. 9. Confirm that `webaccess` connects and exposes 15 tools. Do not add `command`, `args`, environment variables, authentication headers, a Podman path, or a container name to the `webaccess` entry. When LM Studio and Podman run on the same computer, `127.0.0.1` is the correct address. ## Test in LM Studio Load the tool-capable model and ask: > Use `web_search` to find the official LM Studio MCP documentation and return > the source URL. LM Studio should display a `web_search` tool call. ## Start Again Later Start Podman Desktop and its Podman machine. Then open **PowerShell (Standard user)** in the repository root and run: ```powershell .\Start-WebAccessMcp.cmd ``` The command preserves the existing secrets and PostgreSQL page cache. ## Stop the Stack Use **PowerShell (Standard user)** from the repository root: ```powershell Set-Location ".\deploy" podman compose --project-name webaccess-mcp stop ``` ## Troubleshooting If startup fails, run from the repository root using **PowerShell (Standard user)**: ```powershell Set-Location ".\deploy" podman compose --project-name webaccess-mcp ps podman compose --project-name webaccess-mcp logs --tail 200 ``` Detailed guides: - `QUICKSTART.md` - `docs\Standing-Up-the-Podman-Compose-Stack.md` - `docs\Adding-WebAccess-MCP-to-LM-Studio.md` - `docs\Running-in-Podman-Desktop-and-LM-Studio.md`