athena
Deterministic Git & AST change intelligence engine and Model Context Protocol (MCP) server. Detects breaking changes, maps caller blast radius, identifies doc drift, and drafts evidence-grounded release notes with optional local LLM acceleration.
Install / Use
claude mcp add aaryanrwt -- npx -y github:aaryanrwt/athenaIf the server publishes to npm under a different name, use that package instead — check the repo README.
MCP Server
Model Context Protocol server
Quality Score
Category
AI & Machine LearningSupported Platforms
Our assessment of athena
athena scores 75/100 on our quality scale, 845th of 960 AI & Machine Learning skills we index.
Its MCP Server is 40 KB long, well organised into 42 sections with 14 code examples: long enough that it reads more like full documentation than a focused instruction file, which agents can find harder to follow.
It has 10 GitHub stars, so there is little community track record yet; judge it on its content.
Maintenance, license and trust
- The repository was last updated 5 days ago, so athena is actively maintained.
- It is released under the Apache-2.0 license, a permissive license that allows use, modification and commercial use with attribution.
- Its trust signals score 97/100, with no cautions. These come from repository metadata, not a code audit — read the skill file before letting an agent act on it.
athena compared with similar skills
All 4 of these similar skills score higher than athena; compare them before choosing.
| Skill | Score | Stars | Updated | Format |
|---|---|---|---|---|
| athena (this skill)by aaryanrwt | 75 | 10 | 5d ago | MCP Server |
| claude-memby thedotmack | 100 | 103.2k | today | CLAUDE.md |
| Agent-Reachby Panniantong | 100 | 102.6k | 1d ago | CLAUDE.md |
| Understand-Anythingby Egonex-AI | 100 | 88.3k | 3d ago | CLAUDE.md |
| headroomby headroomlabs-ai | 100 | 77.8k | 1d ago | CLAUDE.md |
Frequently asked questions
- How do I install athena?
- Run
claude mcp add aaryanrwt -- npx -y github:aaryanrwt/athena. The install tabs above show the steps for each supported agent. - Which AI agents does athena work with?
- It is written for Claude Code and Claude Desktop, as a MCP Server file. Other agents that read the same format can often use it too.
- Is athena safe to use?
- It is Apache-2.0-licensed and scores 97/100 on trust signals. Skills are instructions an agent will follow, so read the file before installing it and do not approve commands you do not understand.
- Is athena still maintained?
- The repository was last updated 5 days ago, so athena is actively maintained.
Skill content
View source on GitHubAthena (athena-mcp) — v1.1.0
Local-First, Evidence-Backed Multi-Language Git & AST Release-Readiness and Change-Intelligence MCP Server
- Version:
1.1.0 - Owner & Maintainer: Aaryan Rawat (
aaryan28rwt@gmail.com)
Athena is an open-source, local-first Model Context Protocol (MCP) server and developer CLI that delivers verifiable, ground-truth code-change intelligence. By combining deterministic Git diff analysis (100+ commit range inspection), Python Abstract Syntax Tree (ast) and multi-language structural signature comparison (TypeScript, JavaScript, Python, C++, C, Go, Rust, Java, Kotlin, Scala, C#, Ruby, PHP, Swift), downstream caller blast-radius tracing, and documentation-drift detection, Athena eliminates the hallucination and superficiality of conventional release-note generators.
Key Differentiators (v1.1.0)
- Deterministic-First Core (100% LLM-Independent): Athena works out-of-the-box completely offline without Ollama, without an external API, and without any network connection. All Git inspections, AST symbol deltas, caller blast-radius graphs, doc-drift warnings, and release notes execute deterministically with 0ms LLM wait time.
- Multi-Language Structural & AST Diffing (New in v1.1.0): Full Python
astcontract and recursion analysis plus structural symbol & signature comparison across TypeScript, JavaScript, C++, C, Go, Rust, Java, Kotlin, Scala, C#, Ruby, PHP, and Swift — detecting removed public symbols, parameter additions/removals, visibility narrowing, and return-type mutations across 100+ commit histories. - Ground-Truth Evidence Citations: Every reported change story and potential breaking change is bound to verified
EvidenceRecorditems containing the 40-character commit SHA, commit subject, file path, exact base/target line spans, unified diff hunk header/excerpt, and confidence rating. - AST Contract Differencing & Recursive Function Defect Analysis: Athena ignores misleading Conventional Commit prefixes (e.g.
fix:commit messages that actually introduce new public methods or break signatures). It analyzes parameter defaults, positional-parameter index shifts (BUG-030),None-fallback body guards (griffe#469), return-type annotations,(str, Enum)<->StrEnummigrations (griffe#501), Enum members, public constants, newly raised exceptions (griffe#498), and recursive function bugs (broken recursive self-call arity, unconditional infinite recursion without a base case, unqualified class-method self-recursionNameErrors, and self-recursive caller blast radius). - Public API Exposure Differentiation: Automatically distinguishes symbols exported via
__all__, package__init__.pyre-exports, TypeScriptexportdeclarations, Go exported identifiers, Rustpubitems, and Java/C#/C++publicsymbols (is_public_api_export=True) from internal/private helpers (is_public_api_export=False) so teams know when external package consumers are at risk even if internal callers look clean. - Unresolved Dynamic Dispatch Detection & Coverage Confidence Score: Scans repository ASTs for reflection and dynamic invocation patterns (
getattr,hasattr,setattr,globals(),locals(),eval,exec,importlib.import_module, and TypeScript bracket dispatchobj[key]()) and emits an explicitAnalysisCoverageReport(static_parse_coverage_pct,unresolved_dynamic_calls_count,overall_blast_radius_confidence) so CI and PR reviewers never get a false sense of security. - Three Resilient Operating Modes (Supporting
gemma:latest/ Gemma 7B &qwen2.5-coder:7b):- Mode A (Deterministic-only, Default): High-speed, 100% offline rule-based engine.
- Mode B (Optional Local Ollama Enhancement): When requested (
--llm,use_llm=True, orATHENA_ENABLE_LLM=1), local Ollama (gemma:latest/gemma:7borqwen2.5-coder:7b) enhances narrative explanations while strictly verifying evidence citations against deterministic ground truth. Includes automatic VRAM swap eviction for 6 GB GPUs. - Mode C (Fault-Tolerant Fallback): If Ollama times out, returns malformed JSON, or is offline, Athena gracefully degrades to deterministic output with clear diagnostic warnings.
- Drop-In GitHub Actions PR Audit Bot (
.github/workflows/athena-audit.yml): Ships with a ready-to-use GitHub Actions workflow that runs the multi-Python test matrix and posts a structured Breaking-Change, Blast-Radius & Release Notes summary table directly to$GITHUB_STEP_SUMMARYon every Pull Request. - Private, Read-Only & Fully Annotated MCP Tools: Athena declares explicit MCP
ToolAnnotations(readOnlyHint=True,destructiveHint=False,idempotentHint=True,openWorldHint=False) on all 7 registered tools, never modifies your repository, never pushes code, and runs 100% locally on consumer hardware (verified on an NVIDIA RTX 4050 Laptop GPU with 6 GB VRAM and 16 GB RAM).
Architecture
flowchart TD
subgraph Client["MCP Clients & CLI"]
CLI["Athena CLI v1.1.0\n(athena / python -m athena.cli)"]
IDE["MCP Clients\n(Claude Desktop, Cursor, VS Code)"]
end
subgraph Server["Athena MCP Server (stdio)"]
MCP["athena.mcp_server\n(7 Read-Only Annotated Tools & Schema Validation)"]
SEC["Security & Redaction Gate\n- Path Jail & Symlink Defense\n- CLI Flag Injection Defense\n- Bearer / Key Redaction"]
end
subgraph Pipeline["Athena Deterministic Analysis Pipeline (Mode A)"]
GIT["Git Inspector\n(100+ Commit Range, Partial Clone Batching, Diffs)"]
AST["Multi-Language AST & Structural Analyzer\n- Python ast Contract & Recursion Differ\n- TS/JS/C++/C/Go/Rust/Java/C#/Ruby/PHP Scanner\n- Public API & None-Guard Analysis"]
BLAST["Blast-Radius & Caller Engine\n(Self-Recursion, Static Import, Arity & Dynamic Call Tracing)"]
DRIFT["Doc-Drift Scanner\n(Markdown Stale Symbol & Call-Arity Extraction)"]
VERIFY["Evidence & Coverage Verifier\n(SHA, Line Span, Parse Coverage & Confidence)"]
end
subgraph OptionalAI["Optional Local LLM Enhancement (Mode B / Mode C)"]
OLLAMA["Local Ollama Service\n(gemma:latest 7B / qwen2.5-coder:7b)"]
FALLBACK["Graceful Degrader\n(Timeout / Malformed JSON / Offline)"]
end
subgraph Output["Evidence-Backed Results"]
REPORT["Audience-Tailored Release Notes\n(Developers, Maintainers, Clients)"]
BREAK["Potential Breaking Changes\n(Recursion Bugs, Public Exposure, Callers)"]
end
CLI --> SEC
IDE --> MCP
MCP --> SEC
SEC --> GIT
GIT --> AST
AST --> BLAST
BLAST --> DRIFT
DRIFT --> VERIFY
VERIFY -->|Default: Mode A| Output
VERIFY -->|Optional --llm: Mode B| OLLAMA
OLLAMA -->|Enhanced Output| Output
OLLAMA -.->|Failure / Timeout| FALLBACK
FALLBACK -->|Degraded Mode C| Output
Installation & Setup
Prerequisites
- Python:
>= 3.12(verified on Python3.12,3.13, and3.14) - Git:
>= 2.30(verified ongit 2.53) - (Optional) Ollama: Local Ollama server at
http://127.0.0.1:11434withgemma:latest(Gemma 7B,a72c7f4d0a15, ~5.0 GB) and/orqwen2.5-coder:7b(dae161e27b0e, ~4.7 GB) if Mode B is desired.
Portable Installation
Clone the repository and install Athena in a clean virtual environment:
# 1. Clone repository
git clone https://github.com/aaryanrwt/athena.git
cd athena
# 2. Create and activate virtual environment
python -m venv .venv
# On Linux / macOS:
source .venv/bin/activate
# On Windows (PowerShell):
.venv\Scripts\Activate.ps1
# 3. Install in editable development mode
pip install -e .
MCP Server Configuration
Athena communicates via standard JSON-RPC 2.0 over standard I/O (stdio).
Claude Desktop (claude_desktop_config.json)
Add Athena to your Claude Desktop configuration file:
- macOS:
~/Library/Application Support/Claude/claude_desktop_config.json - Windows:
%APPDATA%\Claude\claude_desktop_config.json
{
"mcpServers": {
"athena": {
"command": "athena",
"args": ["mcp"],
"env": {
"ATHENA_ENABLE_LLM": "0"
}
}
}
}
Note: To enable optional local Ollama enhancement (gemma:latest or qwen2.5-coder:7b), set "ATHENA_ENABLE_LLM": "1".
Cursor / VS Code / Other MCP Clients
For generic MCP clients using executable commands:
{
"athena": {
"command": "python",
"args": ["-m", "athena.cli", "mcp"]
}
}
Registered MCP Tools
Athena exposes 7 read-only, schema-validated MCP tools, each annotated with explicit MCP ToolAnnotations (readOnlyHint=True, destructiveHint=False, idempotentHint=True, openWorldHint=False):
| MCP Tool Name | Primary Parameters | Description |
| :--- | :--- | :--- |
| inspect_repository | repo_path (str), max_recent_commits (int, default 10) | Safely inspects branch status, tags, HEAD commit, clean status, and recent commits. |
| analyze_git_history | repo_path (str), base_ref (str, default "AUTO"), target_ref (str, default "HEAD"), max_commits (int, default 100, up to 500), use_llm (bool, default False) | Extracts bounded revision range (100+ commits), commit history, unified diffs, and multi-language AST/structural symbol deltas into grounded change stories. |
| detect_potential_breaking_changes | repo_path (str), base_ref (str, default "AUTO"), target_ref (str, default "HEAD"), include_low_confidence (bool, default False), use_llm (bool, default False) | Detects signature mutations, removed symbols, newly raised exceptions, base class changes, affected downstream callers, and documentation drift. |
| generate_release_notes | repo_path (str), base_ref (str, default "AUTO"), target_ref (str, default "HEAD"), release_version (str, default "Unreleased"), audience (str: end_users | developers | maintainers | clients), output_file (str, optional), overwrite (bool, default False), use_llm (bool, default False) | Synthesizes audience-tailored Markdown release notes with evidence citations, upgrade considerations, and potential breaking changes. |
| analyze_change_graph | repo_path (str), base_ref (str), head_ref (str, default "HEAD"), max_commits (int, default 100) | Inspects deterministic symbol delta graph, caller blast radius, and commit history across base_ref..head_ref. |
| verify_breaking_and_doc_drift | repo_path (str), base_ref (str), head_ref (str, default "HEAD"), include_low_confidence (bool, default False) | Verifies breaking changes, documentation drift, and caller impact across base_ref..head_ref. |
| generate_verified_release_brief | repo_path (str), base_ref (str), head_ref (str, default "HEAD"), release_version (str), audience (str) | Generates structured Markdown + JSON release brief backed by verified commit and diff evidence. |
CLI Usage & Verified Execution Output
The Athena CLI provides instant, zero-LLM deterministic analysis directly from your terminal. All examples below are executed directly against examples/demo_repo, a realistic Git repository with tags v0.1.0 and v0.2.0 (generated via python examples/create_demo_repo.py).
To run from source:
# 1. Generate the local demo Git repository (for testing):
python examples/create_demo_repo.py
# 2. Run Athena CLI (if installed via pip / editable):
athena <subcommand> [options]
# Or directly with Python:
pyt
Truncated for display — read the full file on GitHub.
Related Skills
claude-mem
103.2kPersistent Context Across Sessions for Every Agent – Captures everything your agent does during sessions, compresses it with AI, and injects relevant context back into future sessions. Works with Claude Code, OpenClaw, Codex, Gemini, Hermes, Copilot, OpenCode + More
Agent-Reach
102.6kGive your AI agent eyes to see the entire internet. Read & search Twitter, Reddit, YouTube, GitHub, Bilibili, XiaoHongShu — one CLI, zero API fees.
Understand-Anything
88.3kGraphs that teach > graphs that impress. Turn any code into an interactive knowledge graph you can explore, search, and ask questions about. Works with Claude Code, Codex, Cursor, Copilot, Gemini CLI, and more.
headroom
77.8kCompress tool outputs, logs, files, and RAG chunks before they reach the LLM. 20% fewer tokens for coding agents, 60-95% fewer tokens for JSON, same answers. Library, proxy, MCP server.
Languages
Trust signals
From repository metadata: license, adoption, age and documentation. Not a code audit — see the Safety scan above for what the skill file itself contains.
