AgnesAI Opens Agnes Harness to Deploy AI Agents in Customer Environments
AgnesAI Labs open-sourced a pluggable TypeScript runtime that lets agents share tools, tasks, and interfaces across CLI, Web, and physical devices.
- AgnesAI Labs open-sourced Agnes Harness, a TypeScript agent runtime under Apache-2.0.
- Backend App Server keeps task state shared across CLI, Web, and SDK sessions.
- Plugin system covers backend tools, frontend panels, and full-stack bundles via Cordis.
- Trust model exposes tool approvals, execution constraints, and auditable session records.
- Planned Model Hardware Standard (MHS) will extend the harness to robots and field devices.
- Pre-alpha developer preview; external pull requests are not currently accepted.
AgnesAI Labs has released its agent runtime, Agnes Harness, on GitHub under the Apache 2.0 license. The developer preview targets Forward Deployed Engineering, where engineers work inside a customer’s environment to connect AI systems with operational workflows. It ships as a TypeScript monorepo that developers build from source.
The project consolidates models, tools, task state, and business interfaces behind CLI, web, SDK, and plugin surfaces. Shared infrastructure reduces the glue code required across customer deployments. The repository had more than 1,100 GitHub stars at publication, although its pre-alpha status means developers should expect breaking API changes.
One session, three surfaces
Agnes stores task state in a backend App Server shared by the CLI, web workbench, and SDK. A developer can start a task in the terminal, inspect it in a browser panel, and resume it through the SDK without recreating session logic for each interface.
Cordis manages plugin dependencies and lifecycles, while separate extension paths support backend tools and frontend panels. Services can connect through the Model Context Protocol, or MCP, a standard interface for exposing tools and context to AI applications.
This story is for Pro members
You've reached the end of the free preview. Upgrade to AlphaSignal Pro to read the full article - and everything else behind the paywall.
Comments
