Skip to main content
Glama
jospnunes

Idencify

README.md
# Idencify MCP Server

**Whole-repo code intelligence for AI agents, served over [MCP](https://modelcontextprotocol.io).**

AI coding agents are great at the file they're looking at and blind to the other 2,000. [Idencify](https://idencify.com) gives them the whole-repo picture: connect a GitHub repository and it builds a cross-language call graph with semantic search — served through 14 MCP tools to Claude Code, Cursor, and any MCP client.

> This is the public documentation repository for the hosted server at `https://mcp.idencify.com/mcp`. Sign up at [idencify.com](https://idencify.com) — free tier included.

## Quickstart (Claude Code)

```bash
claude mcp add --transport http idencify https://mcp.idencify.com/mcp
```

Authentication is OAuth 2.1 — your MCP client opens the browser, you approve, done. API-key access (REST) is also available from the dashboard.

## Tools

| Tool | What it answers |
|---|---|
| `analyze_impact` | "What breaks if I change this symbol?" — transitive blast radius |
| `expand_call_graph` | Callers/callees N levels deep, cross-file and cross-language |
| `get_definition` | Definition + dependency context for a symbol |
| `get_context` | Assembled, ranked context pack for a task |
| `symbol_search` | Hybrid semantic + keyword + identifier search |
| `find_pattern` | Structural pattern occurrences across the repo |
| `detect_cycles` | Dependency cycles, ranked by severity |
| `check_boundaries` | Architecture boundary violations |
| `health_score` | Repo/module health grade with hotspots |
| `graph_diff` | Exact graph diff between two commits (adds/removes/breaking) |
| `semantic_changelog` | Human-readable changelog from graph changes |
| `search_history` | Query indexed git history |
| `summarize_file` | File summary with symbols and relations |
| `rate_context` | Feedback loop — rate retrieved context to improve ranking |

## Language support

Rust, TypeScript, JavaScript, Python, Go, Java and C++ — all indexed into one unified graph, with enhanced compiler-level precision for Rust, TypeScript, Python and Go. Re-indexing never loses relations: the graph can only get better over time.

## How it works

1. Connect a repository (GitHub App with read-only access, or REST API for public repos).
2. Idencify indexes it into a whole-repo graph: symbols, calls, imports, containment and implementation relations, plus semantic search — updated automatically on every push.
3. Your agent queries the graph over MCP — every answer is backed by the indexed graph, not guesses.

## Links

- Website / signup: **[idencify.com](https://idencify.com)**
- Dashboard: [app.idencify.com](https://app.idencify.com)
- MCP endpoint: `https://mcp.idencify.com/mcp` (Streamable HTTP; SSE at `/mcp/sse`)
- Contact: hello@idencify.com

## Pricing

Free ($0) · Plus ($9/mo) · Pro ($29/mo) — see [idencify.com](https://idencify.com) for limits.

TDQS

B3.4/5.0

Scored across 15 tools

Disambiguation5/5

Each tool targets a distinct aspect of code analysis—search, definition, context, impact, graph traversal, cycle detection, boundary checks, pattern finding, file summary, health score, graph diff, changelog, history, feedback, and sync. Even seemingly related tools like analyze_impact and expand_call_graph are clearly separated by their focus on reverse dependencies versus call graph walking.

Naming Consistency3/5

The majority of tools follow a verb_noun pattern (get_definition, analyze_impact, detect_cycles), but several deviate—symbol_search, health_score, graph_diff, and semantic_changelog use reversed or noun-first conventions, weakening predictability across the set.

Tool Count5/5

With 15 tools, the server is at the upper edge of the ideal range, yet each tool serves a distinct and necessary function for a comprehensive code intelligence platform, ensuring no tool feels redundant or extraneous.

Completeness5/5

The tool set covers a full lifecycle: ingesting/triggering syncs, searching and retrieving symbols, analyzing dependencies and cycles, enforcing architecture, finding patterns, summarizing files, assessing health, tracking changes over time, and even incorporating feedback—leaving no obvious gaps for its domain.

Maintenance

ActivityMaintained
ResponsivenessNo issues