ContextPro
ContextPro is a local MCP server that helps coding agents find compiler-backed references, search/read code efficiently, assemble task context, and persist project memory.
Find references and usage:
exact_referencesuses TS/JS compiler bindings to find callers/references while excluding unrelated same-named symbols;blast_radiusgives textual usage candidates by kind.Search code:
searchsupports semantic/fuzzy discovery by meaning across files and identifiers;grepsupports literal or regex string search with globs and context limits.Read just enough source:
skeletonshows file API surfaces/signatures without full bodies;explainextracts specific symbols with dependencies and token budgets.Assemble task context:
task_packreturns ranked files, nearby tests, recent diffs, memory hits, risks, and graph-related files for a task.Validate changes:
validate_scoperecommends relevant validation checks and commands after scope selection.Index and inspect projects:
statusgives file counts, index state, memory stats, git branch, uncommitted changes, and recent commits;reindexupdates the project index after pulls/branch switches/external edits.Persist and reuse memory/artifacts:
memory_record_taskstores task outcomes with evidence, risks, assumptions, and source hashes;artifact_readreloads previously generated context artifacts.Discover tools and report efficiency:
discover_toolsfinds/enables expert tools and can run supported read/index tools if the client cannot refresh;efficiency_reportreports auditable context-compression estimates.
Provides compiler-exact references and bounded source context for JavaScript repositories, using TypeScript compiler analysis to distinguish bindings and report reference kinds and truncation.
Supports Python reference lookup through a configured Pyright LSP, with unconfigured Python heuristics retaining distinct provenance.
Provides compiler-exact references and bounded source context for TypeScript repositories, resolving aliases, re-exports, and same-named symbols via the TypeScript compiler.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@ContextPro find all compiler-exact references to parseConfig across this repo"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
ContextPro
Exact find-all-references for your coding agent on TypeScript/JavaScript repos — answered by the TypeScript compiler, not grep.
ContextPro is a local MCP server for agents such as Claude Code, Codex, Cursor and OpenCode. Its compiler-backed "find all references" means a rename or refactor is far less likely to miss callers that import the function under another name or through a re-export (within the files the compiler can check). It also serves source reads with a size limit, so the agent gets the lines it needs instead of whole files. It runs on your machine and needs no API key.
Quick start
Requirements: Node.js 20–26 (24 LTS recommended). No API key or Ollama needed. For compiler references, install the target project's dependencies first (npm install); otherwise files can't be checked.
Fastest try (Claude Code):
cd your-project
claude mcp add contextpro -e CONTEXTPRO_TOOL_PROFILE=lean -- npx -y contextpro-mcp
claude mcp list # contextpro: ... ✔ ConnectedKeep the server name before -e, which would otherwise swallow it.
Claude Code, whole team (commits .mcp.json plus agent instructions in AGENTS.md and CLAUDE.md):
npx -y contextpro-mcp init claude
claude # start it in the repository root and approve "contextpro" onceCodex:
codex mcp add contextpro --env CONTEXTPRO_TOOL_PROFILE=lean -- npx -y contextpro-mcpCodex starts the server in the folder where you start codex, so start it in your project.
Cursor: run npx -y contextpro-mcp init cursor in your project. It writes .cursor/mcp.json with the workspace folder as the root.
Other clients that read mcpServers JSON (in the project's own config file):
{ "mcpServers": { "contextpro": {
"command": "npx",
"args": ["-y", "contextpro-mcp"],
"env": { "CONTEXTPRO_TOOL_PROFILE": "lean" } } } }Never put a fixed --root in a global config: every project would then use that one repository.
OpenCode: run npx -y contextpro-mcp init opencode, then opencode mcp list.
Try it. Ask your agent:
"Use contextpro to find every caller of
<function>in<file>.""Use contextpro to show me how
<feature>works."
INSTALL.md covers VS Code, team approval, a smoke test that works without any client, running from source and troubleshooting.
Related MCP server: Integra CodeMemory
What it is good at
Safe renames and refactors. The compiler tells apart aliased imports, re-exports and different functions that share a name, and the answer says how many results exist and whether any were cut off.
Reading just enough source. The agent asks for a file, a symbol or a line range under a size budget and is told what was left out.
Several agents on one repository. Each Git worktree gets its own index, kept up to date as files change.
A small tool list. The recommended
leansetup shows the agent three tools (context,exact_referencesanddiscover_tools); the other 65 stay hidden until the agent asks for them.
Python references work through a configured Pyright; other languages fall back to structural and text search.
Honest results
ContextPro has not been shown to make agents finish tasks faster. Our own trials, published in docs/benchmarks, include the losses.
It is worth trying for renames and refactors where aliased or re-exported callers matter, or when you don't know where the code lives.
Measured | Result |
Compiler reference checks on two fixtures (barrel aliases, same-named bindings) | 24 of 24 timed queries passed |
Response size on six retrieval tasks vs a scripted | about 41% fewer response tokens, but about 6.8× slower per call |
Matched agent trials (2 read-only tasks), lean profile vs native tools | 23.8% more total input, 19.2% more wall time |
Matched agent trials (2 read-only tasks), workflow profile vs native tools | 128.7% more total input |
For simple lookups, plain grep is cheaper. We welcome reproducible comparisons, including ones where another tool wins; see how we compare.
Learn more
Agent guide: how agents should call the tools, with examples
Configuration: tool profiles, providers and environment variables
Security model: what runs locally and what a hosted provider would see
License and origins
ContextPro began as a fork of Context+ by ForLoop and has since been largely rewritten. It is distributed under the MIT license; the original ForLoop copyright notice is preserved, see NOTICE.
Available Tools
14 toolsartifact_readARead-onlyIdempotent
Load a previously generated context artifact by ID so agents can reuse earlier task packets, repo maps, and decisions without regenerating them.
Inputs: artifact_id (required), root.
blast_radiusARead-onlyIdempotent
Find textual usage candidates with counts by kind. Use kind:call for call-containing lines. Includes same-name symbols; use exact_references for binding-resolved TS/JS rename checks.
Inputs: symbol_name (required), kind, mode, root, +1 more.
discover_toolsB
Discover or enable experts and return their schemas. If the client cannot refresh its catalog, call runs a supported read/index tool; never source edits.
Inputs: call, root, limit, query, +1 more.
efficiency_reportBRead-onlyIdempotent
Report ContextPro's persistent, auditable context-compression estimates by tool. History stores numeric metrics only—never prompts, source code, response bodies, or file paths. Token savings are estimates, not billing data.
Inputs: root, scope, format, history_limit, +2 more.
exact_referencesARead-onlyIdempotent
Find who calls or references a symbol. TS/JS compiler bindings exclude same-named unrelated symbols; results label calls/imports and give per-file call totals before limits. Check scope, ambiguity and warnings. Python is heuristic; imported SCIP freshness is unverified.
Inputs: symbol_name (required), file, line, root, +2 more.
explainARead-onlyIdempotent
Extract a specific function/class body with its dependencies, without reading the entire file. Provide symbol name to get its code + called functions. Provide file path for a structural overview.
Inputs: file, root, depth, format, +3 more.
grepARead-onlyIdempotent
Fast literal or regex string search across the project's indexed file set. Use this for log-message lookups, config-key searches, error-message tracing, or any time you know the exact characters you want to find. Use 'search' instead for semantic intent queries. Skips binary files; respects .gitignore via the indexed walker.
Inputs: pattern (required), glob, root, regex, +7 more.
memory_record_taskB
Persist useful task outcomes with linked files, assumptions, evidence, and open risks for future sessions.
Inputs: task (required), summary (required), root, files, +7 more.
reindexA
Re-walk, re-parse, and re-index project files into the database. Detects changed files via content hash and only re-processes what changed. Eagerly populates file and identifier embeddings unless no_embed=true or CONTEXTPRO_REINDEX_NO_EMBED is set. Use after git pull, branch switch, or external edits. Builds the import graph for related/test_map tools.
Inputs: root, force, no_embed, auto_hubs, +1 more.
searchARead-onlyIdempotent
Search the codebase by MEANING. Modes: 'file' searches file headers/symbols, 'identifier' searches function/class definitions with call chains, 'hybrid' (default) runs both and returns combined results. Uses embeddings for semantic matching plus keyword overlap scoring. For an exact known symbol name in a language with a real language server available (e.g. via an LSP-backed MCP like Serena), prefer that for go-to-definition — this tool is best for meaning-based / fuzzy discovery across the whole repo.
Inputs: query (required), mode, root, top_k, +15 more.
skeletonARead-onlyIdempotent
Get detailed function signatures, class methods, and type definitions of a specific file WITHOUT reading the full body. Shows the API surface and line ranges. Signatures are normalized summaries, not exact edit anchors; use context(targets:[{file,start,end}]) for source text before editing.
Inputs: file_path (required), root, exported_only.
statusARead-onlyIdempotent
Project overview dashboard. Returns file counts, index state, memory graph stats, git branch, uncommitted changes, and recent commits. One call for complete situational awareness at the start of a session.
Inputs: root, format.
task_packARead-onlyIdempotent
Assemble the minimum useful working set for a task: ranked files, nearby tests, recent diffs, memory, and risks.
Inputs: task (required), root, format, preset, +9 more.
validate_scopeARead-onlyIdempotent
After edit_scope, recommend the most relevant validation checks and commands. Pass files when you already know what changed; they take precedence over task inference.
Inputs: task (required), root, files, format, +1 more.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 14 tools
Most tools have clearly distinct roles: grep vs search vs blast_radius vs exact_references are explicitly differentiated by literal/semantic/binding behavior, and status/reindex/task_pack cover separate lifecycle stages. There is still some conceptual overlap among skeleton, explain, artifact_read, and task_pack as ways to retrieve context, but the descriptions give enough guidance to select between them.
All names use snake_case casing, but the semantic pattern is mixed: some are verbs (search, grep, reindex, explain), some are nouns (skeleton, status, task_pack), and others are noun_verb or verb_noun compounds (artifact_read, memory_record_task). The names remain readable, but there is no single predictable verb_noun convention throughout.
At 14 tools, the set is well within a reasonable range for a code-context and analysis server. Each tool appears to serve a distinct workflow: indexing, search, symbol analysis, task packaging, memory, validation, and reporting.
The surface covers index lifecycle, semantic and literal search, reference analysis, symbol extraction, task context assembly, and memory persistence, which is broad for the domain. There are minor gaps: no explicit artifact write/list operation, no direct source-reading/context tool despite being referenced in descriptions, and no separate memory recall/list tool beyond task_pack.
Maintenance
Related MCP Connectors
Source-backed prompting and coding-agent guidance with provenance and freshness checks.
Stateless TS/JS compiler facts for agents: references, imports, impact. No repo index or OAuth.
The context platform for AI agents: code search, team decisions, and grounded project context.
Retrieve citation-ready technical context and coordinate evidence-backed work between AI agents.
Related MCP Servers
- AlicenseNot gradedqualityBmaintenanceProvides structured semantic context for JavaScript/TypeScript codebases, enabling coding agents to navigate, review, and change code with explicit uncertainty.11 npm1MIT
- AlicenseNot gradedqualityBmaintenanceEnables coding agents to explore TypeScript, JavaScript, and Prisma projects through local symbol, call, and model-usage graphs, while recalling persistent project-scoped decisions and returning focused source context.1MIT
- FlicenseNot gradedqualityBmaintenanceEnables AI coding agents to interrogate a local repository deterministically for symbols, dependencies, call paths, routes, tests, and Git impact, returning verified structured evidence with exact line citations instead of requiring the whole codebase to be read.79 PyPI4-
- AlicenseNot gradedqualityAmaintenanceEnables AI coding agents to retrieve bounded, focused context from a local structural code graph, including source entities, Git relationships, and recorded project memory. It supports exploring structure, inspecting callers, callees, and impact, and carrying prior mistakes into later sessions without reading full files.24 npmMIT