Sense
Give Claude Code or Cursor grounded codebase maps—symbols, callers, and blast radius—before refactors or feature work.
Overview
Sense is an MCP server for the Build phase that gives AI coding agents a symbol graph, blast-radius insight, and semantic search over your codebase.
What is this MCP server?
- Builds a symbol graph for navigation across functions, types, and modules
- Blast-radius analysis before you edit shared utilities or APIs
- Semantic search over the repo when grep and filename search fall short
- Ships as stdio MCP bundles (mcpb) for darwin arm64/amd64 and linux arm64/amd64
- Registry version 0.84.3 with Model Context Protocol server.schema.json 2025-12-11
- 4 mcpb platform packages (darwin arm64/amd64, linux arm64/amd64)
- stdio transport
Community signal: 8 GitHub stars.
What problem does it solve?
Agents guess at call graphs and edit the wrong files because they only see partial context from open tabs and naive search.
Who is it for?
Indie devs mid-build on a monorepo or multi-package app who refactor often with Claude Code, Cursor, or Codex.
Skip if: Greenfield repos with a handful of files where reading every file in chat is still faster than indexing.
What do I get? / Deliverables
After you register Sense over stdio, the agent can answer structural questions and narrow risky edits using graph-backed codebase understanding.
- Registered Sense MCP server in your agent host config
- Agent-callable codebase understanding (symbol graph, blast radius, semantic search) over your project
Recommended MCP Servers
Journey fit
Solo builders wire agent tooling while implementing; Sense is installed when the product codebase is real enough to need structural answers, not during idea sketches. Agent-tooling is the shelf for MCP servers that extend how coding agents read and reason about your repo, distinct from shipping a user-facing frontend.
How it compares
Repo intelligence MCP server, not a test runner, linter skill, or deployment integration.
Common Questions / FAQ
Who is Sense for?
Sense is for solo builders and small teams using AI coding agents who need symbol-level navigation, impact awareness, and semantic search across a growing codebase.
When should I use Sense?
Use Sense during build and agent-tooling setup when refactors, cross-module features, or unfamiliar directories outgrow simple grep and partial file reads.
How do I add Sense to my agent?
Add the Sense MCP server to your host config using stdio and the mcpb release for your OS and CPU (darwin or linux, arm64 or amd64) from the v0.84.3 GitHub release, then restart the client.