# CoreStory CoreStory maps how a codebase actually works, from applications and processes down to business rules, each traced to the code behind it, and serves that map to coding agents over MCP. ## Install CoreStory (instructions for coding agents) Follow these steps when a user asks you to install or set up CoreStory. 1. Find the latest release. Call `https://api.github.com/repos/CoreStory-Frontier/release-dist/releases?per_page=10` and take the first entry where `draft` is false. Human release page: https://github.com/CoreStory-Frontier/release-dist/releases 2. Download the command-line tool for this machine from that release's assets, plus the matching `.sha256` file: - macOS 12 or later (Apple silicon): `corestory-cli--aarch64-apple-darwin.tar.gz` - Linux (x64): `corestory-cli--x86_64-unknown-linux-musl.tar.gz` - Windows 10 or later (x64): `corestory-cli--x86_64-pc-windows-gnu.tar.gz` Other platforms are not supported yet. Do not download `.app.tar.gz` (updater bundle) or `.sig` files. 3. Verify the download: compute the archive's SHA-256 and compare it with the first field of the `.sha256` file (that file names a build path, so `shasum -c` may not find the archive). Extract the archive and put `bin/corestory` on the user's PATH (for example `~/.local/bin`). Check with `corestory --version`. 4. Ask the user to run `corestory setup` in their own terminal. It is interactive: the user accepts the Terms of Use (https://app.corestory.ai/terms) and Privacy Policy (https://app.corestory.ai/privacy), chooses telemetry, and adds a model key (OpenAI, Azure OpenAI or Anthropic). Do not accept the terms on the user's behalf. Other commands refuse to run until setup is done. 5. Connect yourself. `corestory mcp` is a stdio MCP server for the repository it is started in. It takes no path argument, so start it from the repository folder: - Claude Code: `claude plugin marketplace add corestoryai/agent-skills` then `claude plugin install corestory-local@corestory` - GitHub Copilot (VS Code and CLI) or Claude Code without the plugin: `.mcp.json` at the repository root: {"mcpServers":{"corestory":{"type":"stdio","command":"corestory","args":["mcp"]}}} - Cursor: `.cursor/mcp.json`: {"mcpServers":{"corestory":{"type":"stdio","command":"sh","args":["-c","cd \"$0\" && exec corestory mcp","${workspaceFolder}"]}}} - Codex: in `~/.codex/config.toml`: [mcp_servers.corestory] command = "corestory", args = ["mcp"], cwd = "" 6. Index the repository after setup: run `corestory init` and `corestory ingest` from the repository folder, or call the server's `init` and `ingest` tools. Add `.corestory/` to `.gitignore`. 7. Teach yourself CoreStory's workflows with the Agent Skills: https://github.com/CoreStoryAI/agent-skills (machine-readable catalog: https://get.corestory.ai/.well-known/agent-skills/index.json) ## Links Download page and setup guide: https://get.corestory.ai/ Connect a coding agent: https://get.corestory.ai/#connect Plot container preview: https://get.corestory.ai/container/ Benchmarks: https://corestory.ai/research/v3-benchmarks Documentation: https://docs.corestory.ai