
Kastra
- 1 repo stars
- Updated July 23, 2026
- kastra-labs/claude-plugin
Sign up for Kastra and govern your AI agents — onboarding tools + edge install.
About
kastra is a Claude Code skill in the AI & Agent Building category. Sign up for Kastra and govern your AI agents — onboarding tools + edge install.
- kastra
- AI & Agent Building
- AI-coding skill
Kastra by the numbers
- Data as of Jul 24, 2026 (Skillselion catalog sync)
/plugin marketplace add kastra-labs/claude-plugin/plugin install kastra@kastraAdd your badge
Show developers this skill is listed on Skillselion. Paste this into your README.
| repo stars | ★ 1 |
|---|---|
| Last updated | July 23, 2026 |
| Repository | kastra-labs/claude-plugin ↗ |
What it does
Sign up for Kastra and govern your AI agents — onboarding tools + edge install.
README.md
Kastra Claude plugin
Govern your AI agents with Kastra. This plugin lets your assistant inspect your Kastra governance state read-only — policies, decisions, the tamper-evident audit chain, kill-switch incidents, and environments — plus create a free account and install local enforcement.
Add Kastra to Claude in one click (claude.ai)
Opens the "add custom connector" dialog in claude.ai pre-filled with Kastra's
hosted MCP (https://api.kastra.ai/mcp) — no copy-pasting a URL. It gives your
assistant the onboarding tools (kastra_start_signup, kastra_signup_status)
so you can spin up a free-tier account without leaving the chat.
Claude Code plugin
/plugin marketplace add kastra-labs/claude-plugin
/plugin install kastra@kastra
What you get
Read-only governance inspection via the Kastra OAuth connector
(https://api.kastra.ai/mcp/account). The first time a governance tool runs,
Claude Code opens your browser to authorize once (OAuth 2.1 + PKCE, scope
kastra.account.read) — it never sees a pasted API key. Then your assistant can
read your policies, decisions, audit chain, incidents, and activity stats.
Governance slash commands (composed from the read-only tools):
| Command | What it does |
|---|---|
/kastra:governance-status |
Snapshot: decision volume, deny rate, latency, active incidents, per-environment enforcement/shadow state |
/kastra:policy-check |
Active policy for an environment, which rules are firing, and recent governance changes |
/kastra:decision-audit |
Search decision history and explain denials ("why was I blocked?") |
/kastra:verify-audit |
Verify the tamper-evident decision audit chain is intact |
Onboarding tools (kastra_start_signup, kastra_signup_status) backed by
the hosted MCP — create a free-tier account without leaving the chat. They create
no account and send no email themselves; they return a link you open to finish a
normal, email-confirmed signup.
/kastra:edge-install skill — guides setup of the local enforcement edge
(kastra-edge) so your agent's tool calls are actually blocked/held against your
policies. See below.
Inspect vs. enforce
This plugin is read-only inspection. To make Kastra actually enforce (block
or hold tool calls) on this machine, install the local edge — run
/kastra:edge-install or brew install kastra-labs/tap/kastra-edge && kastra-edge login && kastra-edge install-claude. Local enforcement runs as a Claude Code
hook from the kastra-edge CLI; bundling it directly into this plugin is a
planned future version.
Note: the governance connector registers as an MCP server named
kastra, so its tools surface asmcp__kastra__*. If you also have the localkastra-edgeMCP installed under the same name, expect a naming overlap.