Back to directory

flutter_agent_harness

A cross-platform AI agent harness for Dart and Flutter with streaming provider adapters, native tool calling, session persistence and context compaction, running on Flutter apps, the Dart VM and the web.

flutter_agent_harness

flutter_agent_harness is a pure-Dart agent core, with no dependency on dart:io or Flutter in its library code, so the same agent loop compiles and runs on the Dart VM, Flutter desktop and mobile apps, and the web, with platform-specific capabilities kept behind an execution-environment abstraction. Provider failures never throw; every error, including network failures, rate limits and malformed streaming data, arrives as a typed error event so the agent loop keeps running instead of dying on a dropped connection.

It supports each provider's native tool-calling format, including OpenAI's tools, Anthropic's tool use and Google's function calling, with prompt-based tool calling available only as an opt-in adapter for chat-only runtimes, plus token-based context accounting, model-role fallback chains, API-key rotation and cancellation support throughout. Sessions are stored as append-only JSONL trees supporting branching and labeled navigation, and the project ships a terminal coding agent built on the same core, with a web demo, an iOS beta and a macOS app.

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.