Comments
>log in to comment- PDFops· 3mo ago
The "without framework sprawl" framing resonates — the frameworks that hurt are the ones that own the control flow, so you end up fighting the abstraction the moment an agent does something the happy path didn't anticipate. Where do you draw the harness/framework line — is it that the harness stays a library you call, rather than a runtime that calls you?
ThinHarness is a minimal, opinionated Python harness for building purpose-built AI agents.
- for
- Python developers building bounded agent loops inside their own applications.
- pricing
- open source
- license
- MIT
Key features
- Scoped filesystem tools — Provides built-in tools for safe file system access within agent loops.
- Parallel LLM calls — Fans out LLM requests for efficient parallel processing or majority-vote reliability.
- Structured output & retries — Handles tool output formatting and automatic retries on failure.
- Human approvals & subagents — Supports human-in-the-loop approvals and spawning of sub-agents.
- Search tools — Includes ripgrep-based document search and a JSONL search tool for structured corpora.
- OpenTelemetry tracing — Emits tracing data for observability of agent execution.
Use cases
- Automating business document workflows with fast search and structured tool calls.
- Running parallel LLM evaluations to improve reliability of results.
- Building internal assistant agents that need human approval steps.
- Creating sub-agent hierarchies for complex multi-step tasks.
ThinHarness vs alternatives
| Best for | Small, forkable agent harness | Feature-rich LLM app frameworks | Multi-agent platforms | Data-centric LLM apps | TypeScript AI agents |
|---|---|---|---|---|---|
| Pricing | Open source | Free | Subscription | Free | Free |
| DevHunt upvotes | 3 | 0 | 0 | 0 | 0 |
| Launched | Jun 2026 | — | — | — | — |
- ThinHarness vs LangChain: LangChain offers a larger, more feature-rich framework with many integrations, while ThinHarness stays minimal.
- ThinHarness vs CrewAI: CrewAI focuses on multi-agent orchestration and platform services; ThinHarness provides only the core loop.
- ThinHarness vs LlamaIndex: LlamaIndex centers on data indexing and retrieval; ThinHarness emphasizes agent loop primitives.
- ThinHarness vs Mastra: Mastra is a TypeScript framework, whereas ThinHarness is a Python-only minimal harness.
ThinHarness FAQ
What language does ThinHarness support?+
ThinHarness is written for Python and requires Python 3.11 or newer.
Is there a deployment layer included?+
No, ThinHarness only provides the agent loop; serving, auth, and storage are left to the host application.
Can I use Bash commands in agents?+
Bash is excluded by default; an optional BashTool can be enabled for prototyping.
How large is the codebase?+
The repository contains about 8,035 lines of code across 24 files.
Is ThinHarness open source?+
Yes, it is MIT-licensed and pre-1.0 open source.
Summarized by DevHunt from ryanbbrown.com · Oct 4, 2026. Details may change; check the official site.