B — Some risk signals Graded October 16, 2026Minor concerns worth a quick check.Based on public adoption and transparency signals as of October 16, 2026. Our opinion, not a code audit or security guarantee.

App connector · Developer tools, Memory & knowledge

code-review-graph

Local-first code intelligence graph for MCP and CLI. Builds a persistent map of your codebase so AI coding tools read only what matters, with benchmarked context reductions on reviews and large-repo workflows.

Bottom line: Signal-based grade (4/7 points): very widely starred on GitHub, public source code. This reflects adoption and transparency, not a code audit.

Popularity score: 50/100
Rank: #116 of 500
GitHub stars: 32.2k

What it does

code-review-graph builds a persistent, local-first intelligence graph of your codebase that AI tools can query to understand code structure and dependencies. It reduces context window usage for code review and large-repository workflows by filtering to only relevant code files, available as both an MCP server and CLI tool.

Who it’s for

Engineering leaders and development teams managing large codebases who want to optimize AI-assisted code review and reduce token costs. Builders integrating AI coding assistants into existing workflows.

Copy-paste setup

Check these steps against the project’s own instructions before running anything.

Verify against the repo README at https://github.com/tirth8205/code-review-graph (opens external site) for the latest installation method. Expected patterns:

CLI (typical):

bash

npm install -g code-review-graph
code-review-graph init /path/to/repo

MCP Server (Claude Desktop): Add to claude_desktop_config.json:

json

{
  "mcpServers": {
    "code-review-graph": {
      "command": "npx",
      "args": ["code-review-graph", "server"]
    }
  }
}

Risks

Data exfiltration: Graph construction indexes entire codebase locally; verify it doesn't transmit code or dependency metadata to external services.
Credentials in analysis: If the tool scans environment files or config, sensitive credentials could be indexed into the persistent graph and exposed to connected AI tools.
Prompt injection via graph queries: Malicious code comments or file paths could inject instructions into AI prompts if graph output is passed unsanitized.
Supply-chain dependency risk: Tool depends on underlying graph/AST libraries; audit transitive npm dependencies for known vulnerabilities.
Persistence security: Local graph file should be excluded from version control and protected with file-system ACLs; no encryption mentioned in signals.

How to stay secure

Least-privilege permissions

Run code-review-graph as a non-root user with read-only access to source code; write access limited to a single isolated graph storage directory.
Restrict file-system ACLs on the persistent graph file (default location) to 0600 (owner read/write only); ensure it's not world-readable or group-accessible.
If using Docker or containers, mount codebase as read-only volume and grant write access only to a dedicated graph output path outside version control.
Audit and disable any indexing of .env, secrets/, credentials/, or similar sensitive paths via configuration allowlist/blocklist.

Network isolation

Verify code-review-graph operates entirely offline; disable or block any HTTP/HTTPS calls to external services during code indexing and graph construction.
If the tool or its dependencies attempt outbound connections, block at firewall/network policy level and log attempts; confirm in network monitoring before trust.
Run in an air-gapped environment or container with no external connectivity if analyzing sensitive/proprietary codebases.
Review all npm transitive dependencies for any telemetry, analytics, or phoning-home behavior; use npm audit and snyk to detect high-risk packages.

Monitoring

Log all graph construction invocations, file paths scanned, and any errors or warnings; alert on unexpected file access patterns or large data reads.
Monitor the persistent graph file for unauthorized modifications (file-integrity monitoring, e.g., aide, tripwire, or OS-level auditing).
Audit connections to AI tools or MCP agents that query the graph; ensure only authorized tools receive graph output and log all queries.
Set up alerts for graph file growth anomalies or repeated re-indexing of excluded paths, indicating potential misconfiguration or abuse.

Red flags to watch for

Graph file contains plaintext secrets, API keys, passwords, or internal URLs; any indexed .env, config, or credentials files should trigger immediate quarantine.
Presence of outbound DNS queries, HTTP requests, or file exfiltration during indexing; verify tool is truly local-first with packet inspection or process monitoring.
Code comments or file paths in the graph appear altered or contain suspicious prompt-injection patterns (e.g., ); sanitize graph output before MCP relay.
Transitive npm dependency vulnerabilities (especially AST/parsing libraries used for graph construction); run npm audit --audit-level=moderate and remediate high/critical findings before deployment.

Alternatives

Codebase.ai (opens external site) – Cloud-based code indexing and search for AI agents (different trust model).
Repo2Vec (opens external site) – Embedding-based repository understanding (focus on semantic similarity).
Sourcegraph (opens external site) – Enterprise code intelligence platform with AI integration (centralized, not local-first).
Dust (opens external site) – Workflow platform with built-in code retrieval (managed service alternative).

Sources

github.com/tirth8205/code-review-graph