pacifio/cersei

What it solves

Cersei is a Rust SDK designed to simplify the creation of production-ready coding agents. It provides the necessary infrastructure—such as tool execution, LLM streaming, memory management, and sub-agent orchestration—as composable library functions, allowing developers to build embeddable agents or CLI tools like the included "Abstract" CLI.

How it works

Cersei operates as a modular framework where an Agent is configured via a builder pattern. It abstracts LLM providers (Anthropic, OpenAI, and others) and integrates a wide array of built-in tools for filesystem access, shell execution, and web searching. It employs a three-tier memory system (flat files, CLAUDE.md hierarchy, and an optional graph-based memory layer via Grafeo) to maintain context and session persistence.

Who it’s for

It is intended for Rust developers who want to build high-performance, low-latency coding agents or replacements for existing CLI agents (like Claude Code) without having to implement the core agentic loop and tool-dispatch logic from scratch.

Highlights

• High Performance: Extremely low tool dispatch and startup latency compared to Node.js-based alternatives.
• Extensive Toolset: Over 30 built-in tools for coding, planning, and orchestration.
• Graph Memory: Optional graph-backed memory for near-instant recall without requiring additional LLM calls.
• Sub-Agent Orchestration: Ability to spawn parallel workers and coordinate tasks via a coordinator mode.
• MCP Integration: Support for the Model Context Protocol (MCP) to connect to external servers.
• Flexible Provider Support: Native support for Anthropic (including OAuth) and OpenAI-compatible APIs (Ollama, Azure, vLLM).