Release 0.5 — execution — is out

Install. Run. Extend.

OpenClerq is a persistent local administrative agent — like a local administrative worker. It keeps context, tools, skills, and configuration on your machine and exposes them through a desktop app plus a small gateway.

Technically, it's a local model context server (MCS) for administrative and clerical work. All calculations run locally; AI provides guidance and explanations only. Skills are available by module — build your own or subscribe to hosted modules.

Since 0.4 the gateway is authenticated and bound to loopback: every endpoint but the health check requires a bearer token generated on first run, and there is no development bypass. Since 0.5 it talks to eight model providers — six in the cloud under your own keys, or local models through Ollama and LM Studio.

Version
0.5.0
License
MIT
Platforms
macOS · Windows
Models
8 providers
Calculations
Local
Gateway
127.0.0.1, authed

Who it's for

Anyone who wants to run a local administrative agent — accountants, agencies, HR teams, legal assistants, small business, self-hosters, and developers. No cloud required; your data stays on your machine.

Vision: Replace repetitive clerical work by 2030 — AI handles routine, rule-based office work; humans stay in control of decisions and filings.

How to get it

Installers

Prebuilt macOS (.dmg) and Windows (.msi/.exe) on GitHub Releases. Install, launch — gateway and calculation engine start automatically.

Hosted

For commercial Clerq (scalable MCS, teams, custom modules), see Clerq for Business.

Latest release — 0.5, execution

Release 0.5 connects OpenClerq to the models you already pay for and puts every call on the record: which run made it, what it cost, and how to stop it. One configuration change: the undocumented CLERQ_OPENAI_API_KEY alias is gone — set OPENAI_API_KEY instead.

  • Eight model providers — Anthropic, OpenAI, DeepSeek, Moonshot, Z.ai and MiniMax in the cloud, or Ollama and LM Studio on your own machine. The registry is a file you can edit.
  • Chat console — talk to any configured model raw, run it through the full pipeline, or send one message to up to eight models and compare latency, tokens and cost.
  • Every run on the record — tasks, triggers, webhooks and chat messages are recorded with a step trace and their cost. Any run can be cancelled, and a kill switch stops everything at once.
  • Private by default — state lives in a local SQLite store, ~/.clerq is readable only by you, and the vault key moves into the OS keychain.

The sandbox, policy engine and git workspaces ship in 0.5 as tested libraries; the agent runtime in 0.6 is the first thing to use them. See the roadmap, or read the changelog in full.

Trust and clarity

Open source — MIT license Authenticated, loopback-only gateway; vault key in the OS keychain Fork heritage — inspired by OpenClaw Use at your own risk — see NOTICE

Guides, comparisons and use cases

Plain answers for offices adopting AI, and honest comparisons with the tools you already know — including where they are the better choice.

For businesses

Need a specific desktop worker built on OpenClerq — HR, legal, admin, marketing, or something else? Tell us what you need and we'll design and price a module for you.

Clerq for Business