Claude Code plugins from Tessary AI — drop-in tooling for teams building with LLMs.
In any Claude Code session, add this marketplace and install a plugin:
/plugin marketplace add tessaryai/plugins
/plugin install <plugin-name>@tessary
To update later:
/plugin marketplace update tessary
- evals — generate a calibrated eval suite for your LLM product.
- crew — a multi-agent dev harness that triages, implements, reviews, and maintains your repo.
Point evals at your repo and it generates a complete eval suite for your LLM features: one grader per failure mode (judge prompt, rubric, self-tests), plus a visual report you can open in your browser. Have production traces? Hand them in and graders get calibrated against real data.
/plugin install evals@tessary
Then run /evals:synthesize-graders in a Claude Code session, or just ask Claude to "synthesize evals for this repo."
See the plugin README for flags, packs, and validator usage.
crew is a multi-agent development harness. Hand it a goal and an orchestrator decides what
needs doing and does it — triaging issues, implementing changes through a deliberative team,
reviewing PRs, responding to review feedback, and keeping docs and knowledge current —
running unattended up to a review-ready PR (it never merges; you do). You can also invoke
any single capability directly.
/plugin install crew@tessary
Then either let it drive — /crew:run "close out the open bugs" — or call a primitive
yourself, e.g. /crew:review-pr 42 or /crew:triage-bug 17. Run /crew:init-config once to
tune it to your repo. This release is local-first (runs in your own Claude Code session);
GitHub Actions packaging is planned.
See the plugin README for the full skill list and configuration.
MIT — see LICENSE.