--- title: "Getting started with Cork's MCP tools: get_symbol, blast_radius, recap, remember · Cork" description: "Use Cork's MCP tools in Claude Code, Codex and Cursor: get_symbol, blast_radius, recap, remember and recall, with when to call each and example prompts." canonical: "https://corkcli.com/blog/cork-mcp-tools-get-symbol-blast-radius-recap-remember" last_updated: "2026-11-14" published: "2026-10-03" author: "Kevin Vera" --- # Getting started with Cork's MCP tools: get_symbol, blast_radius, recap, remember Use Cork's MCP tools in Claude Code, Codex and Cursor: get_symbol, blast_radius, recap, remember and recall, with when to call each and example prompts. By Kevin Vera · October 3, 2026 · [Guides](https://corkcli.com/blog/c/guides) · 6 min read Cork's MCP tools give a coding agent facts about the code and about other agents' work when it needs more than its briefing: `get_symbol` for one symbol, `blast_radius` for what a change would affect, `recap` for what happened and what it cost, and `remember` and `recall` for durable notes anchored to code. If you installed Cork with `curl -fsSL https://corkcli.com/install.sh | sh`, the tools are already registered for Claude Code, Codex and Cursor. Any other MCP client can use Cork's MCP server at `https://mcp.corkcli.com`. ## A quick refresher on MCP The Model Context Protocol (MCP) lets an agent call tools that live outside it. Each client keeps its own list of servers: - **Claude Code:** `claude mcp add --transport http ` adds a remote server. The default scope is local to the project; `--scope project` writes a shared `.mcp.json` you can commit, and `--scope user` makes it available in all your projects. `claude mcp list` shows what is configured, and `/mcp` inside a session shows status and handles sign-in for servers that need it. - **Codex:** HTTP servers go in `config.toml` under `[mcp_servers.]` with a `url`. `codex mcp login ` authenticates an OAuth server. The Codex CLI, IDE extension and ChatGPT desktop app share this configuration. - **Cursor:** servers go in `.cursor/mcp.json` in the project or `~/.cursor/mcp.json` globally, with a `url` for remote servers. You do not need any of this for the three tools above if you used Cork's installer, which registers Cork's MCP tools for Claude Code, Codex and Cursor. It is useful if you work in another MCP client, or want to check what is configured. ## Where the tools fit Most of Cork's work happens without tool calls. Its hooks hand each agent a briefing when it starts or gets a prompt (the exact signatures, callers and recent changes of the code it is about to touch), tell it about other agents' changes on its next step, and gate writes. The MCP tools are for the cases the briefing does not cover: a symbol outside the current task, a question about impact, a look back, or a decision worth keeping. A good rule: if the briefing already answered it, do not call a tool. Each call is a step the agent spends. ## get_symbol: the facts about one symbol Use `get_symbol` when an agent needs the facts about a specific function, class or method that was not in its briefing. Example prompts: > Before you use `parse_config`, get_symbol it so you have the current signature. > Use get_symbol on `billing/invoice.py::total_due` and summarize what you learn before editing it. This is cheaper than having the agent open and search several files to reconstruct the same facts. ## blast_radius: what a change would touch Use `blast_radius` before a change that others depend on: a rename, a new required argument, a changed return type. It tells the agent what depends on the code it is about to change. Example prompts: > Run blast_radius on `api/client.py::fetch_user` before renaming it, and list everything we would need to update. > I want to make `tenant_id` required in `create_order`. Check the [blast radius](https://corkcli.com/blog/blast-radius) first. On a shared repo, this pairs well with a message. If the blast radius includes code another agent is working on, tell them before you change it: ```bash cork message --about api/client.py::fetch_user "Renaming to get_user in the next few minutes" ``` After the change, Cork's [semantic lint](https://corkcli.com/blog/semantic-lint) turns the rename or signature change into a rule other agents are checked against, and they hear about it on their next step. ## recap: what happened and what it cost Use `recap` to look back: what agents did, and what it cost. It is a quick answer to "what changed while I was away?" before you dig into commits. Example prompts: > Give me a recap of what happened in this repo this morning. > Recap the work on the payments module and what it cost. On the paid plan, history covers 90 days; Cloud Free keeps a 30-day window; Enterprise retention is custom. ## remember and recall: notes anchored to code Use `remember` when an agent settles a decision that should outlive the session: a constraint, a gotcha, a reason something looks odd. The note is anchored to code, so it comes back to agents working on that code through `recall`. Example prompts: > Remember on `cache/user_cache.py::get` that we must not cache admin users; permissions change without invalidation. > Before you touch the retry logic, recall what we decided about it. Good notes are short and specific: what to do or not do, and why. Notes that only restate the code waste everyone's context. On the paid plan, Cork adds shared knowledge across the team. ## A short workflow that uses all four 1. **Start a task.** Read the briefing. If it mentions a function you need but does not describe, `get_symbol` it. 2. **Before changing a shared interface,** run `blast_radius`, and message whoever is working on affected code with `cork message --about`. 3. **When you learn something non-obvious,** `remember` it on the code it applies to. 4. **When you come back to the repo,** start with `recap`, then `recall` on the area you are about to work in. ## How Cork helps Cork is the [coordination layer](https://corkcli.com/blog/coordination-layer-ai-coding-agents) for multiplayer coding agents: it keeps Claude Code, Codex and Cursor agents, on one machine or across a team's machines, from overwriting each other on one codebase, and hands each agent what the others already learned. The MCP tools are how an agent asks for that knowledge directly, on top of the briefings, live awareness, [write gate](https://corkcli.com/blog/write-gate), semantic lint and landing that the hooks provide with no extra model calls. The install registers the tools for Claude Code, Codex and Cursor: ```bash curl -fsSL https://corkcli.com/install.sh | sh ``` ## Frequently asked questions ### Do I need to add Cork's MCP server by hand? Not for Claude Code, Codex or Cursor: the installer registers Cork's MCP tools for all three. For any other MCP client, add the server at `https://mcp.corkcli.com` using that client's MCP configuration. ### When should an agent call get_symbol instead of reading the file? When it needs the facts about one specific symbol that its briefing did not include. Asking directly can save the agent from searching and reading several files to piece them together. ### What is the difference between remember and a comment in the code? A comment lives in the file and gets read only when someone opens it. A note made with `remember` is anchored to the code and comes back to agents through `recall` when they work there, and it can record reasons that do not belong in the source. ### Does blast_radius work across teammates' work? `blast_radius` reports what depends on the code you are about to change. On the paid plan, agents across the team see each other, so combine it with `cork message --about` to reach whoever is working on the affected code. ## Sources - - - - ## Related reading - [MCP servers for coordinating coding agents: what they are and how to use one](https://corkcli.com/blog/mcp-servers-coordinating-coding-agents): What an MCP server for coordinating coding agents does, how it differs from hooks, and how to connect one to Claude Code, Codex, Cursor and Gemini CLI. - [Messaging between your coding agents with cork message](https://corkcli.com/blog/cork-message-coding-agents): Use cork message to reach every coding agent in a repo, one agent, or whoever is working on a function, across Claude Code, Codex and Cursor. With examples. - [Why AI agents re-read the same files, and how to stop it](https://corkcli.com/blog/why-ai-agents-reread-same-files): Why Claude Code, Codex and Cursor agents keep re-reading the same files, and how to stop it: specific prompts, lean context, symbol lookups and briefings. - [Cork Free: set up multiplayer coding agents on one machine in one command](https://corkcli.com/blog/cork-free-setup-one-machine): Set up Cork Free in one command so Claude Code, Codex and Cursor agents on one machine stop overwriting each other. No account; data stays on the machine. [All articles](https://corkcli.com/blog.md) · [Cork in Markdown](https://corkcli.com/index.md) · [llms.txt](https://corkcli.com/llms.txt) This page in HTML: https://corkcli.com/blog/cork-mcp-tools-get-symbol-blast-radius-recap-remember