What is GliderMCP?

GliderMCP makes local MCP servers for coding agents. Glider gives an agent C#/.NET navigation, diagnostics, analysis, edits, and refactors. Scout searches repository files across languages. GliderTrace runs .NET commands and collects runtime evidence. Keep stores shared memory that agents can save to and search. Each product runs on your computer. See the documentation.

Are the products free to use? Can I use them at work?

Every product is free for non-commercial use and work on open-source projects. Your organization can evaluate them at work for 30 days. Commercial use of Glider and GliderTrace after the evaluation requires a subscription. Commercial use of Scout and Keep stays free of charge for now. See Pricing.

Which product do I need?

Use Glider for C#/.NET. Scout searches repository files across languages. GliderTrace executes commands and collects .NET runtime evidence. Keep gives several agents and conversations one memory for decisions and notes. You can connect products independently; see all products. Check Glider compatibility for .NET Framework requirements.

Which clients work with the products?

Scout requires a client that can launch a local stdio server. Glider and GliderTrace support stdio or Streamable HTTP. Keep runs as one token-protected HTTP server that every client connects to. Choose your client in the installation guide for Glider, Scout, GliderTrace, or Keep.

For HTTP, workspace paths, worktrees, multiple instances, and connection recovery, see setup options. For rejected arguments, see troubleshooting. Choose a model that supports your client’s MCP tools.

Should several agents share one server?

For Glider and GliderTrace, use one localhost HTTP instance per workspace and connect each agent client to that endpoint. Clients share workspace state; coordinate code edits and runtime tasks. Use separate instances and ports for different worktrees or permissions.

Scout supports stdio only. Assign its tasks to one agent and configure the server only for that agent, where your client supports separate configurations. Task assignment alone cannot reduce processes if every agent starts every server.

Keep is built to be shared: run one Keep server and connect every agent to it. Only one process can open a Keep store at a time.

Duplicate stdio registrations can multiply memory, watchers, and file descriptors. Choose either a plugin or a manual entry for each product. Replace its stdio entry when you select HTTP.

See the product setup guidance for Glider, Scout, or GliderTrace.

Can the agent change my code or run commands?

Glider includes tools that write files and apply refactors. Preview support and defaults differ by tool. GliderTrace can execute commands. Scout keeps source files unchanged. Check the tool reference and your MCP client approval settings.

Does my code leave my machine?

The servers process code and artifacts locally. Keep stores its documents on your computer, in ~/.glider/keep by default. Your MCP client can send returned content to its model provider. Product telemetry and public agent feedback are separate channels. See Privacy for their scope and controls.

How do I buy and activate a subscription?

Create an account on the account page with Google or your email address, then choose a plan. A personal subscription assigns its seat to you. For a team, invite each developer and assign a seat. Each developer then signs in once from Glider or GliderTrace and approves the device in the browser. See Pricing for plans and team seats.

The server says my release is expired or close to expiry. What now?

Future packages will expire 90 days after release. Existing versions retain their original expiry. Update it and restart the client; this is separate from license duration. Follow the update guide for Glider, Scout, or GliderTrace.

How do I search files Glider does not load?

Glider’s search_text covers loaded workspace documents, including declared additional files and analyzer configuration. Use Scout for files outside that scope. See when semantic tools help for tasks that need more than text matches or model inference.