Tetriz Harness launches team-wide agent control
Tetriz Harness launches two MIT-licensed, self-hosted tools for auditing and distributing AI coding-agent rules, skills, hooks, permissions, and MCP configuration across Claude Code, Codex, Cursor, and Kiro. Its scanner grounds recommendations in repository and team signals, while the control plane rolls approved changes out to targeted developers. [Product Hunt](https://www.producthunt.com/products/tetriz-harness) [GitHub](https://github.com/tetriz-ai/tetriz-harness)
Tetriz Harness targets the overlooked infrastructure layer behind agent quality: the configuration that quietly determines how coding agents behave. The concept is compelling, though its early v0.1.0 status and integration requirements make it better suited to technically mature teams today.
- –Detects conflicting rules, missing files, ineffective patterns, and overly broad permissions with evidence.
- –Converts successful team practices into centrally managed harness components instead of relying on tribal knowledge.
- –Supports test/control rollouts, but teams must manually define cohorts and measure outcomes externally.
- –Derived recommendations require populated signal data, an ingestion adapter, and a model key.
- –MIT licensing and self-hosting give engineering teams control over sensitive session and repository data.
DISCOVERED
2h ago
2026-10-16
PUBLISHED
9h ago
2026-10-16
RELEVANCE
AUTHOR
[REDACTED]