codebase-intel

Your AI agent writes code. But does it know why your code exists?

AI coding agents can autocomplete, refactor, and generate code. But they still fail at the things that matter most in production:

• They don't know your team decided to use token bucket over sliding window — and why
• They don't know the compliance team requires rate limit headers on every response
• They don't know that changing config.py will break billing and analytics
• They generate code that looks right but violates your project's architectural patterns

codebase-intel fixes this. It's the context layer that sits between your codebase and any AI agent — providing not just what code exists, but why it exists, what rules it must follow, and what breaks if you change it.

Before vs After

WITHOUT codebase-intel                    WITH codebase-intel
─────────────────────                     ───────────────────

Agent reads: every file in the dir        Agent reads: only what matters
Tokens used: 16,063                       Tokens used: 5,955
Knows why code exists: No                 Knows why: Yes (13 decisions)
Quality guardrails: None                  Guardrails: 4 contracts enforced
Drift awareness: None                     Drift: stale context detected
Impact analysis: None                     Impact: knows what else breaks

Result: faster but fragile                Result: faster AND correct

Real benchmarks on production codebases

Numbers from codebase-intel benchmark on real production repos. The token reduction comes from targeted graph traversal. The decisions and contracts are what no other tool provides.

What makes this different

There are great tools for code graphs (code-review-graph is excellent — 6K+ stars). We don't compete with them. We solve what they don't:

The missing layer

What exists today:          What codebase-intel adds:

Code → Graph → Agent        Code → Graph ──────────────────→ Agent
                                     ↓                         ↑
                              Decision Journal ──→ WHY ────────┤
                              Quality Contracts → RULES ───────┤
                              Drift Detector ──→ WARNINGS ─────┘
                              
                              "Here are the 3 files that matter,
                               the decision your team made 6 months ago,
                               and the 2 rules you must not violate."

Quick Start

pip install codebase-intel

# Initialize on your project
cd your-project
codebase-intel init

# See what it found
codebase-intel status

# Mine decisions from git history
codebase-intel mine --save

# Run benchmarks (see the before/after)
codebase-intel benchmark

# View live dashboard
codebase-intel dashboard

Connect to Claude Code

{
  "mcpServers": {
    "codebase-intel": {
      "command": "codebase-intel",
      "args": ["serve", "/path/to/your/project"]
    }
  }
}

Now your agent automatically gets relevant context, decisions, and contracts before writing code.

The Three Pillars

1. Decision Journal — "Why does this code exist?"

Every team makes hundreds of decisions that never get documented. Why did you choose Postgres over Mongo? Why is auth middleware structured that way? Why was the sliding window approach rejected?

codebase-intel captures these from git history automatically and links them to code:

# .codebase-intel/decisions/DEC-042.yaml
id: DEC-042
title: "Use token bucket for rate limiting"
status: active
context: "Payment endpoint was getting hammered during flash sales"
decision: "Token bucket algorithm with per-user buckets, 100 req/min"
alternatives:
  - name: sliding_window
    rejection_reason: "Memory overhead too high at scale"
constraints:
  - description: "Must not add >2ms p99 latency"
    source: sla
    is_hard: true
code_anchors:
  - "src/middleware/rate_limiter.py:15-82"

Without this: your agent proposes sliding window (the exact approach you rejected 6 months ago).
With this: your agent sees the decision, follows it, and respects the SLA constraint.

2. Quality Contracts — "What rules must AI follow?"

Linters check syntax. Contracts enforce your project's patterns:

# .codebase-intel/contracts/api-rules.yaml
rules:
  - id: no-raw-sql
    name: No raw SQL in API layer
    severity: error
    pattern: "execute\\(.*SELECT|INSERT|UPDATE"
    fix_suggestion: "Use the repository pattern"

  - id: async-everywhere
    name: All I/O must be async
    severity: error
    pattern: "requests\\.(get|post)"
    fix_suggestion: "Use httpx.AsyncClient"

Built-in AI guardrails catch the patterns AI agents mess up most:

• Hallucinated imports (modules that don't exist)
• Over-abstraction (base classes with one subclass)
• Unnecessary error handling for impossible conditions
• Comments that restate code instead of explaining why
• Features that weren't requested (YAGNI violations)

3. Drift Detection — "Is our context still valid?"

Context rots. Decisions get outdated. Code anchors point to deleted files. codebase-intel detects this:

$ codebase-intel drift

╭──────────────── Drift Report ────────────────╮
│ Overall: MEDIUM                               │
│ 3 items need attention                        │
╰───────────────────────────────────────────────╯

- [MEDIUM] Decision DEC-012 anchored to deleted file
- [MEDIUM] Decision DEC-008 is past its review date
- [LOW] 2 files changed since last graph index

19 Languages

Full tree-sitter parsing via tree-sitter-language-pack:

CLI Commands

codebase-intel init [path]              # Initialize — build graph, create configs
codebase-intel analyze [--incremental]  # Rebuild or update the code graph
codebase-intel mine [--save]            # Mine git history for decision candidates
codebase-intel drift                    # Run drift detection
codebase-intel benchmark                # Measure token efficiency (before/after)
codebase-intel dashboard                # Live efficiency tracking over time
codebase-intel serve                    # Start MCP server
codebase-intel status                   # Component health check

MCP Tools (7 tools)

Community Contract Packs

Pre-built quality rules for popular frameworks:

cp community-contracts/fastapi.yaml .codebase-intel/contracts/

Architecture

AI Agent (any) ──→ MCP Server (7 tools)
                        │
                  Context Orchestrator
                  (token budgeting, priority, conflicts)
                   ╱          │          ╲
          Code Graph    Decision     Quality
          (19 langs)    Journal      Contracts
          SQLite+WAL    YAML files   YAML+builtins
                   ╲          │          ╱
                    Drift Detector
                    (staleness, rot, orphans)
                          │
                    Analytics Tracker
                    (live efficiency metrics)

The Philosophy

We don't make AI agents smarter. We make them informed.

An agent with 1M tokens of context is like a developer with access to every file in the company — overwhelming and unfocused. An agent with codebase-intel is like a developer who just had a 5-minute conversation with the senior engineer: "Here's what you need to know, here's why we did it this way, and here are the three things you absolutely cannot break."

We don't compete with graph tools. We complete them.

Code graphs answer "what depends on what." That's necessary but not sufficient. codebase-intel answers the harder questions: Why was this decision made? What constraints apply? What will the compliance team flag? What did we already try and reject?

We don't hide the truth. We prove it.

Run codebase-intel benchmark on your project. See the numbers. Run codebase-intel dashboard over time. Watch the improvement. Every claim is backed by reproducible, project-specific data.

Contributing

Areas with the most impact:

1. Contract packs — Share quality rules for your framework
2. Language extraction — Improve parsing for specific languages
3. Decision mining — Better git history analysis
4. Benchmarks — Run against your repos, share results

git clone https://github.com/MutharasuArchunan13/codebase-intel.git
cd codebase-intel
pip install -e ".[dev]"
pytest

License

MIT