# Mingbird

> Updated 2026-09-30 · type: tool · category: mcp-agents · status: active · rev 1

Mingbird gets small 2–9B local models to finish real multi-step tasks on a laptop, running offline against a local Ollama backend.

- Open source: yes (Apache-2.0)
- Self-hostable: yes
- Pricing model: free
- Best for: A builder who wants an offline, self-hosted agent that runs on small local models (2–9B via Ollama) on an ordinary laptop without a big GPU or a cloud API — and who believes reliability comes from the harness, not just the model.
- Not for: A team that needs frontier-model reasoning, a managed cloud agent, or production maturity today — Mingbird is brand-new (~1 star) and bets on small-model task completion that depends heavily on the chosen local model.
- Last verified: 2026-09-30

- **Canonical:** https://gtmstacker.com/registry/tool/mingbird/
- **Source:** [Mingbird · GitHub](https://github.com/Mingbird/Mingbird-agent)
- **Tags:** mcp-agents, ai-infrastructure, agent-harness, local-first, self-hostable, small-models
- **Repository:** https://github.com/Mingbird/Mingbird-agent

## Is Mingbird open source?

Yes, Mingbird is open source under the Apache-2.0 license.

## How much does Mingbird cost?

Mingbird is free to use.

## Can I self-host Mingbird?

Yes, Mingbird can be self-hosted (the source is available under the Apache-2.0 license).

## Alternatives & related

- [OpenShell](https://gtmstacker.com/registry/tool/openshell/)
- [Jeff](https://gtmstacker.com/registry/tool/jeff/)
- [Jeeves](https://gtmstacker.com/registry/tool/jeeves/)


---

Mingbird is a local-first agent harness that gets small 2–9B models to finish real multi-step tasks on an ordinary laptop, running offline against a local Ollama backend. Open source: yes (Apache-2.0); self-hostable with a one-click zero-outbound mode and all config in `~/.ollama_agent/config.json`; free OSS. It is brand-new (~1 star, created 2026-09-17) but active — 30 commits, v1.8.2, a 461-test suite, and CI on Linux and macOS.

## What it does

Mingbird is the scaffolding around a small local model, not another model. Its bet is that most agent failures are harness defects — bad tool wiring, lost state, no retries — rather than the model being too small, so it focuses on the loop that turns a 2–9B model running under Ollama into something that actually completes a multi-step task on a normal laptop. You run it fully offline: it talks to a local Ollama backend, offers a one-click zero-outbound mode so nothing leaves the machine, and keeps its configuration in `~/.ollama_agent/config.json`. The project ships a 461-test suite and a 288-cell public benchmark of harness-vs-model behavior. Open source: yes (Apache-2.0); self-hostable; free OSS.

## Provenance

- Apache-2.0 per repo (license read from GitHub metadata + README, not guessed); ~1 star; created 2026-09-17, pushed 2026-09-23; 30 commits; v1.8.2; 461-test suite; CI on Linux/macOS (github.com/Mingbird/Mingbird-agent, verified 2026-09-30).
- Local-first/offline-first: local Ollama backend, one-click zero-outbound mode, config in `~/.ollama_agent/config.json`, no network required.
- Vendor framing: "harness defects, not model defects," backed by the maker's own 288-cell benchmark — a vendor-claim, not independently reproduced here.
- Surfaced in the GTM Stacker studio daily pull (2026-09-30 pass); license/facts independently verified 2026-09-30.

## Why it matters for a GTM stack

Most GTM agent work does not need a frontier model: classify a reply, draft a follow-up, extract fields, decide the next step. Mingbird is a bet that a small model on your own laptop can do that reliably if the harness around it is good — which, if it holds, means an offline, zero-cost, zero-data-egress agent for the routine steps of an outbound or ops workflow. The honest read: this is a brand-new project (~1 star) and the "it's the harness, not the model" thesis is the maker's own, benchmarked only by the maker. It belongs on the watch list as part of the small-local-model lineage (alongside jeff and jeeves for decisions, openshell for isolation) — try it on a low-stakes task first and judge completion quality on your own data before it runs anything that touches a real system.