Harness Rs

crates.io license

Agent = Model + Harness. This is the Harness — the scaffolding that turns an LLM into an autonomous agent. Any domain: research, ops, assistants, data, code.

A Rust framework for production agents. Compile-time type-safe, deterministic-first, observable, governance built in.

• A loop, not a call — ReAct with tool dispatch, sensor feedback and auto-fix, under a budget; guards for stuck detection, acceptance, spill, deadlines.
• Autonomy earned in stages — L1 report → L2 assisted (human gate) → L3 unattended (allowlist only). Graduate the agent as you build trust.
• Honest isolation — macOS Seatbelt and Docker enforce; a git worktree isolates changes, not capability, and reports itself that way.
• Memory that compounds — procedural, semantic and episodic, across sessions.
• Code does what code can — lint · format · git run as Sensors and Hooks, not model turns. Measured, not asserted.

Quick start

use harness_core::Task;
use harness_loop::AgentLoop;
use harness_models::ApiKind;
use harness_tools_fs::{ListDir, ReadFile};
use std::sync::Arc;

// One model API: protocol family + base_url + model + key. No hardcoded URLs.
let model = ApiKind::OpenAI.build("https://api.deepseek.com", "deepseek-chat", key);
let task = Task { description: "What is the workspace name?".into(), source: None, deadline: None };
let outcome = AgentLoop::boxed(model)          // `boxed` takes Arc<dyn Model>
    .with_tool(Arc::new(ReadFile))
    .with_tool(Arc::new(ListDir))
    .run(task, &mut harness_context::default_world("."))
    .await?;

Register tools, skills, guides, sensors and hooks with #[harness::tool] / #[skill] / #[guide] / #[sensor] / #[hook] — they auto-register via inventory. Scaffold a project with harness new; harness code is an approval-gated coding REPL built on the framework (--yolo to unattend).

More

docs/guide.md — crates, composable layers, sandboxing, PII redaction, documents/OCR, telemetry, record/replay, grounded search, examples · docs/benchmarks.md — pass^k, guard ablation, cost · DESIGN.md · CHANGELOG.md · MIT