title: "Cognitum One launches MetaHarness open-source agent factory" slug: "cognitum-one-launches-metaharness-open-source-agent-factory" published: "2026-10-31" beat: "Launches" tags: ["Launches", "Tools"] creator: "Agentry Newsroom" editor: "Susanne Sperling, Editor — Human in the Loop" tools: ["Claude (Anthropic)", "Perplexity Sonar"] creativeWorkStatus: "verified" dateReviewed: "2026-10-31" aiActArticle50: "compliant" humanView: "https://agentry.news/launches/cognitum-one-launches-metaharness-open-source-agent-factory" agentView: "https://agentry.news/agent/cognitum-one-launches-metaharness-open-source-agent-factory"
Cognitum One unveiled MetaHarness on October 28, 2026, at GitHub Universe 2026 in San Francisco—an MIT-licensed open-source tool that generates customizable AI agent frameworks from GitHub repositorie
Drafted by an AI agent. Verified by Susanne Sperling, Editor — Human in the Loop. AI policy.
Cognitum One announced MetaHarness on October 28, 2026, in conjunction with GitHub Universe 2026, held at Fort Mason Center in San Francisco. MetaHarness is positioned as an open-source factory for agent frameworks that can start from a GitHub repository or a blank project and generate a branded agent harness ready for deployment.
The tool is available under the MIT License with a GitHub repository at ruvnet/metaharness and an npm command-line interface. Each generated harness includes an npx command-line tool, a Model Context Protocol server, scoped memory, a governance policy framework, and witness-signed releases—components designed to give enterprises reproducibility and auditability in agent deployments.
The timing at GitHub Universe—Salesforce's flagship developer conference—signals positioning toward the growing ecosystem of builders shipping production agents. MetaHarness targets teams that need to scaffold governance and provenance into agent systems from the start rather than bolting it on afterward. The open-source release model mirrors successful developer tools like LangChain and LlamaIndex, lowering barriers to adoption while building community contribution pathways.
The package is immediately usable: developers can pull the npm tool, specify a project source or start blank, and receive a structured harness with policy templates and signing infrastructure. This removes boilerplate work for teams building multiple agent variants or versioning agent deployments across teams.
A distinguishing feature is the inclusion of witness-signed releases—cryptographic attestation that a built harness matches its source configuration. This addresses a concrete pain point in enterprise deployments: proving to compliance and security teams that a running agent hasn't drifted from its audited state. Scoped memory compartmentalization further isolates agent state, reducing lateral attack surface in multi-agent systems.
The governance policy component provides templating for access control, logging, and audit hooks—pre-built constraints that enterprises require before deploying autonomous systems in production environments.
MetaHarness enters a crowded developer tools space alongside frameworks like Anthropic's Model Context Protocol, Langgraph, and proprietary orchestration platforms. The open-source licensing and GitHub-native distribution suggest Cognitum One is betting on community adoption and corporate pickup through extension rather than direct licensing revenue.
The release is available now via npm and GitHub, with full source code open for review and contribution.