Back to directory

Nanocodex

A headless, library-first Rust SDK for embedding a complete OpenAI coding-agent loop, including retained sessions, typed history, tools, branches and retries, into other products.

Nanocodex

Nanocodex is a library-first SDK that embeds the complete OpenAI Responses agent loop, retained sessions, typed history, tools, branches, events, retries and cleanup, so a caller does not have to rebuild passing previous messages and response IDs on every turn, a separate state machine for prompt ordering and reconnect replay, or a second orchestration runtime when an agent forks or delegates work.

It is deliberately scoped to one supported OpenAI coding-agent stack rather than a general provider abstraction, and is not an app server; a native CLI and TUI, browser apps, Python and JavaScript packages, durable actors, sandboxes, a voice client and an evaluation harness are all separate consumers built on the same library contract. Callers can present the agent's ordered, typed event stream through any interface, such as a terminal interface, xterm.js or React, rather than adopting a fixed protocol.

Its only desktop application is a native SwiftUI and AppKit macOS app with a tiled workspace, a persistent sidebar, streamed conversations and an agent activity menu bar; a separate mobile inbox targets iPhone and iPad. Nanocodex is published to crates.io and distributed under the MIT license.

More in Agent frameworks

See all 902
  • Recurse

    recurse.run

    A serverless harness that lets a coding agent design, test, and deploy its own specialist sub-agents as tools, MCP servers, or bots, defined by a schema-validated manifest.

  • ProofForge

    shulgin.is-a.dev

    An AI-agent pipeline that decomposes a mathematics problem, proves each piece, and formalizes the result in Lean 4 and Mathlib until the Lean kernel accepts it, with six of its proofs already merged into Google DeepMind's formal-conjectures repository.

  • local-coder

    github.com

    Builds a local coding-agent team on top of Ollama and OpenCode by detecting the machine's hardware, recommending models for seven specialist roles, and testing whether each assigned model can actually use tools, edit a repository and pass verification.