MCP.so
Sign In

Aegis

@fuwasegu

Repository

About Aegis

A deterministic context compiler for AI coding agents. Aegis replaces RAG with a Directed Acyclic Graph (DAG) to strictly enforce software architecture and reduce token consumption by up to 12x.

#context#document#cursor#claudecode#codex#knowledge

Config

Add this server to your MCP-compatible client using the configuration below.

{
  "mcpServers": {
    "aegis": {
      "command": "npx",
      "args": [
        "-y",
        "@fuwasegu/aegis",
        "--surface",
        "agent"
      ]
    },
    "aegis-admin": {
      "command": "npx",
      "args": [
        "-y",
        "@fuwasegu/aegis",
        "--surface",
        "admin"
      ]
    }
  }
}

Tools

6

Compile deterministic context for target files. Returns base documents, resolution path, and templates. v2: delivery-aware with budget control.

Record an observation event. Writes to Observation Layer only (never Canonical).

Retrieve details of a past compile_context invocation.

List distinct intent tags from tag_mappings (approved-resolvable only) with knowledge_version and a SHA-256 tag_catalog_hash of the tag list for client-side caching.

Read-only workspace snapshot: recent compile regions (from compile_log), unresolved compile_miss observations without proposals, and pending proposal count. Does not mutate Canonical Knowledge.

Analyze a project and generate an initialization preview. Does not modify Canonical.

Overview

What is Aegis?

Aegis is a deterministic context compiler for AI coding agents (e.g., Cursor, Claude Code, Codex). It replaces probabilistic search with a Directed Acyclic Graph (DAG) backed by SQLite to map file paths and layers directly to Markdown documentation, ensuring agents receive exact architectural context for each editing task.

How to use Aegis?

After installing and configuring the server, agents invoke Aegis before editing a file; it deterministically traverses the dependency graph and compiles the required context. The exact setup and invocation commands are not detailed beyond the paradigm described.

Key features of Aegis

  • Deterministic Context Routing (same input, same context, reproducible).
  • Transitive Dependency Resolution (automatically fetches dependent guidelines).
  • Dual Surface Architecture (Agent read-only context vs. Admin human review).
  • Human-in-the-Loop Knowledge Evolution (proposes DAG updates, approved by humans).
  • Massive Token Efficiency (12x reduction in token consumption, 3.5x faster responses in benchmarks).
  • Optional Local SLM Expansion (node-llama-cpp engine for semantic intent extraction).

Use cases of Aegis

  • AI coding agents needing architectural guidelines for a specific file or layer.
  • Preventing AI hallucinated architectures by feeding deterministic context.
  • Reducing token waste from agents endlessly grepping repositories.
  • Maintaining codebase integrity with a “Tech Lead” that forces rules before edits.
  • Enabling agent self-review and human-approved knowledge base updates.

FAQ from Aegis

How does Aegis differ from traditional RAG?

Traditional RAG suffers from a “semantic gap”—searching by feature keywords fails to retrieve abstract architectural rules. Aegis uses deterministic compilation via a DAG, not probabilistic vector embeddings.

What are the runtime dependencies?

Aegis uses a SQLite-backed DAG. It optionally supports node-llama-cpp (Qwen) for local SLM expansion, but this is opt-in.

Where does the context data live?

All rules and dependency mappings are stored in a SQLite database that backs the DAG.

Can agents update the knowledge base?

Agents can trigger aegis_observe to report missing rules, but canonical knowledge updates require human approval on the admin surface.

What transports/auth are supported?

—

Frequently asked questions

How does Aegis differ from traditional RAG?

Traditional RAG suffers from a “semantic gap”—searching by feature keywords fails to retrieve abstract architectural rules. Aegis uses deterministic compilation via a DAG, not probabilistic vector embeddings.

What are the runtime dependencies?

Aegis uses a SQLite-backed DAG. It optionally supports node-llama-cpp (Qwen) for local SLM expansion, but this is opt-in.

Where does the context data live?

All rules and dependency mappings are stored in a SQLite database that backs the DAG.

Can agents update the knowledge base?

Agents can trigger `aegis_observe` to report missing rules, but canonical knowledge updates require human approval on the admin surface.

Comments

Basic information

Category

Other

Transports

stdio

Author

fuwasegu

Submitted by

Fuwasegu

More Other MCP servers

Nginx UI

0xJacky

Yet another WebUI for Nginx

11.3K

FastMCP v2 🚀

jlowin

🚀 The fast, Pythonic way to build MCP servers and clients.

26.0K

Production-ready MCP integrations for AI applications

Klavis-AI

Klavis AI: MCP integration platforms that let AI agents use tools reliably at any scale

5.8K

🚀 Model Context Protocol (MCP) Curriculum for Beginners

microsoft

This open-source curriculum introduces the fundamentals of Model Context Protocol (MCP) through real-world, cross-language examples in .NET, Java, TypeScript, JavaScript, Rust and Python. Designed for developers, it focuses on practical techniques for building modular, scalable,

16.7K

IDA Pro MCP

mrexodia

AI-powered reverse engineering assistant that bridges IDA Pro with language models through MCP.

9.9K