Open Source

Building the Future of AI Infrastructure

We build open-source tools for developers and businesses. Self-hostable, production-ready, and MIT licensed. No vendor lock-in. Your infrastructure, your rules.

Built in RustMIT LicensedOpen Source Forever

Scroll

Self-hostable, MIT licensed

Open Source

Run our runtime, context memory, tool layer, and search on your own infrastructure.

Runtime

A.R.E.S

Agentic Retrieval Enhanced Server

A production-grade agentic chatbot server built in Rust. Multi-provider LLM support, tool calling, RAG, and advanced research capabilities , all in one self-hostable package.

Multi-Agent Orchestration

Tool Calling

RAG & Knowledge Bases

Streaming Responses

JWT Authentication

Local-First (Ollama)

Quick Start

# Clone and setup

git clone https://github.com/dirmacs/ares

cd ares

cp .env.example .env

# Start Ollama (local LLM)

ollama pull granite4:tiny-h

# Build and run

cargo build

cargo run

# Server runs on http://localhost:3000

Tool layer

Thulp

Execution context engineering

Rust toolkit for AI agents: tool discovery, validation, execution. MCP via UTCP, Query DSL, skill workflows, SKILL.md parsing. Eleven crates, 183 tests. Apache 2.0 / MIT.

Unified tool abstraction

MCP via UTCP

Query DSL, skill workflows

Session management

Install

cargo install thulp-cli

# MCP with Ares:

thulp-mcp = { version = "0.2", features = ["ares"] }

MCP server

Daedra

Web search & research MCP server

DuckDuckGo search and URL-to-Markdown fetch. STDIO + SSE, built-in caching, library and CLI. Used by Ares for built-in web search. No API keys.

search_duckduckgo

visit_page (Markdown)

STDIO + SSE

Caching, connection pooling

Install

cargo install daedra

# Claude Desktop: add daedra to mcpServers

# "command": "daedra", "args": ["serve", "--transport", "stdio"]

Context memory server

Eruka

Eruka is a context memory server that helps your agents remember what truly matters. It tracks what is known, inferred, and still unknown so they can pull the right facts at the right time instead of guessing, especially in long-running and high-stakes workflows.

"Your agents forget important things, just because they can't pull the right context. Let's fix it."

For Developers

Who is ARES for?

Whether you're building your first AI agent or deploying enterprise solutions, ARES provides the infrastructure you need.

Developers

Build AI agents without wrestling with infrastructure. Focus on your application logic while ARES handles the heavy lifting.

Hot-reload config
Type-safe tool calling
OpenAPI docs auto-generated

Self-Hosters

Run everything locally with Ollama and SQLite. No external dependencies, no API keys required to get started.

Local-first by default
SQLite included
Zero config to start

Enterprises

Production-ready security with JWT auth, API keys, and Argon2 password hashing. Deploy on your own infrastructure.

Data sovereignty
Audit ready
Multi-provider flexibility

Our Philosophy

Why Open Source?

We believe the best tools should be accessible to everyone. No gatekeeping, no vendor lock-in.

Community First

MIT licensed. Use it, modify it, sell it , we don't care. Just build cool stuff.

Your Data, Your Rules

Self-host on your infrastructure. Your data never leaves your servers.

Transparency

Every line of code is auditable. No black boxes, no hidden APIs.

Growing Together

Contributions welcome. Report bugs, suggest features, or submit PRs.

Ready to Build Something?

ARES is open source and ready to use. Clone the repo, run the server, and start building. Or reach out if you need help with custom development.

Eruka ARES Thulp Daedra Contact Us