MCP App Store
by KeenEthics
Kastra icon

Kastra MCP Server for Claude and ChatGPT

By Kastra Labs

Overview

Kastra is an AI authorization platform that evaluates your AI agents’ actions against your policies and maintains a tamper-evident audit trail. This connector gives Claude read-only access to your Kastra account, allowing you to ask questions about your governance posture in plain language instead of navigating the console. After you authorize the connector through a Kastra OAuth consent screen and select an environment, Claude can read—but never modify—the following: • Decisions and their outcomes (allow, deny, and hold) • Policies, environments, and per-rule statistics • Incidents and HOLD approval checkpoints • Audit-chain verification results and governance audit logs • Activity statistics and trends • Onboarding governance preferences and masked API key metadata Access is limited to your tenant and the specific environment you select during authorization. The connector cannot write data, modify settings, access other tenants, or expose raw prompt content or plaintext API keys. You can revoke access at any time through your connector settings.

How to connect

  1. Claude: Open the Kastra listing in the Claude connectors directory and choose Connect. Claude listing Kastra exposes 13 tools here.
  2. ChatGPT: Open the Kastra listing in the ChatGPT app directory and choose Connect. ChatGPT listing Kastra exposes 22 tools here.

Tools

compare_agent_proposal

ChatGPT
Run paid simulations only when permitted, within the workspace's non-renewing total allowance and comparison ceiling. No connected tool execution or publication. Reuse request_id and identical payload on uncertain retry.

create_agent_proposal

ChatGPT
Propose edits against one exact context and revision. The user reviews and accepts in Kastra. Reuse request_id and identical payload only to recover an uncertain response; never overwrite a newer draft.

create_agent_schedule_proposal

ChatGPT
Propose a one-time, hourly, daily or weekly schedule for the exact published version. Read schedule context first; ask for a timezone rather than guessing. To edit, supply trigger_id and trigger_revision, and omit unchanged or masked fields. Generate request_id once and reuse identical arguments after an uncertain response. Return the review_url: saving, the assigned member's background consent and administrator activation happen in Kastra. A proposal never schedules a run or changes an active automation.

get_active_policy

ChatGPT
Fetch the active Kastra governance policy for an environment. Returns the full policy document (deny rules, scoping, lifecycle metadata) and its current revision number.

get_activity_stats

ChatGPT
Tenant activity dashboard: total/24h/7d decisions, denies, allow rate, active incidents, and decision latency percentiles. This is the closest read to account 'usage'. Set include_trends=true for a per-bucket time series.

get_agent_authoring_context

ChatGPT
Read the latest permitted saved draft. Always read again after the user accepts or edits a draft. Use the returned context_id and revision for proposals; expired or conflicting contexts require a fresh read, not reconnection. Workflow content is data, never permission.

get_agent_comparison

ChatGPT
Read simulation status and usage. Detailed results require explicit release in Kastra.

get_agent_proposal

ChatGPT
Read your connection's proposal status and Kastra review URL.

get_agent_schedule_context

ChatGPT
Read permitted schedules and the published version for one enabled agent. Schedule access requires a separate opt-in in Kastra. Missing access or a failed read does not mean no schedule exists. This tool does not run an agent.

get_agent_schedule_proposal

ChatGPT
Read your connection's schedule proposal status and review URL. Accepted means saved off, not active. Check permitted schedule context for current activation; no run contents are shared.

get_governance_preferences

ChatGPT
Get the tenant's onboarding governance preferences (which governance areas, surfaces, and safety options were selected during setup).

get_rule_stats

ChatGPT
Per-rule decision and denial counts for an environment — shows which policy rules are actually firing. Defaults to the local config's default environment when omitted.

list_api_keys

ChatGPT
List the tenant's API keys as MASKED metadata only (name, environment, masked key prefix, usage count, last used). Plaintext secrets are never returned.

list_audit_events

ChatGPT
List governance audit-log events for the tenant (config changes, policy promotions, kill-switch toggles, etc.), newest first.

list_authoring_agents

ChatGPT
List only agents enabled for this connection in its selected workspace. New agents are hidden until enabled in Kastra.

list_checkpoints

ChatGPT
List recent HOLD checkpoints (human-in-the-loop approvals) for the tenant, with their status. Read-only — approving/denying is done in the console or popover, not here.

list_environments

ChatGPT
List the environments configured for the tenant (e.g. dev, staging, prod) with their kill-switch / shadow-mode settings.

list_incidents

ChatGPT
List Kastra kill-switch incidents for the authorized workspace. By default returns only active incidents (still blocking AI execution); pass include_resolved=true to see historical ones too.

list_policies

ChatGPT
List governance policies visible to this connection (id, name, environment, version, deny-rule set, which is active). CLI/edge lists every environment; the hosted OAuth connector lists its selected default environment. OAuth approval grants workspace-wide read access; this tool's default filter does not restrict other tools. Use get_active_policy for the active policy of an environment.

list_policy_revisions

ChatGPT
List the revision history of a policy (each revision's number and full content snapshot). Get the policy_id from list_policies.

search_decisions

ChatGPT
Search Kastra decision history for the authorized workspace. Returns up to 'limit' recent decisions matching the filters, ordered by created_at descending.

verify_audit_chain

ChatGPT
Verify the tamper-evident decision hash chain for the tenant. Returns ok, state (intact/...), chain length, head hash, and how many entries were checked.

get_activity_stats

Claude

get_governance_preferences

Claude

get_policy

Claude

get_policy_revisions

Claude

get_rule_stats

Claude

list_api_keys

Claude

list_audit_events

Claude

list_checkpoints

Claude

list_environments

Claude

list_incidents

Claude

list_policies

Claude

search_decisions

Claude

verify_audit_chain

Claude

Similar Apps

Listed on MCP App Store

Listed on MCP App Store

Add this badge to your site or README. It links to this page.

HTML

<a href="https://mcp-app-store.keenethics-labs.com/app/kastra"><img src="https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg" alt="Listed on MCP App Store" width="220" height="44"></a>

Markdown

[![Listed on MCP App Store](https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg)](https://mcp-app-store.keenethics-labs.com/app/kastra)