One airlock connector.
For every AI client.
Link Notion, GitHub, Slack or any MCP / API you can name, once. Set policy, package Agents and Skills, accumulate organization memory, audit every action. Expose to your preferred AI client.
from Cursor · waiting on @maria
verified · attached to 3 agents
Claude Desktop
Remote HTTPAdd airlock as a custom connector by URL.
https://mcp.air-lock.ai/org/your-org-slug- Open Settings → Connectors.
- Click the + next to Connectors, then Add custom connector.
- Paste the URL above and click Add.
- Complete the sign-in in the browser when prompted.
One connector. Every AI client.
Every AI client your team uses talks to airlock: Claude, ChatGPT, Cursor, Copilot, Gemini, Claude Code and many others. Switch or add a client and nothing re-onboards.
- Cross-tool proxy: one airlock connector fronts every MCP-compatible client.
- Cross-vendor skill & agent distribution: ship a Skill or Agent once and every connected client (Claude, ChatGPT, Cursor, Copilot, Gemini, Claude Code) picks it up automatically, even the ones that don't natively support Skill governance.
- Credential vault: zero-trust; agents never see your real API keys.
- No lock-in: change vendors without losing setup, audit history, or memory.
One connector. Tools discovered. Tokens saved.
Watch a real prompt flow through airlock inside Claude Desktop: the + menu connector, just-in-time tool discovery, the token saving versus a full-schema dump, and the approval link before anything runs.
Three honest steps. No hyperbole.
airlock is a connector, not a black box. Here's exactly what you do.
Pick a connector or bring your own.
1.000+ connectors in the catalog. Anything missing? Upload an OpenAPI spec and airlock generates the MCP server. OAuth handled where supported.
Set per-tool policy.
Tag tools Read-only or Destructive. Route writes for approval to Slack, Teams, or the Control Room. Set rate limits per team or key.
Every action lands in the log.
Watch real-time analytics, roll back Skills, kill switch per server, export EU AI Act evidence on demand.
Package your expertise. Ship it everywhere.
Your best agents live as pasted prompts. airlock turns them into versioned, verified objects. Build once, ship to every team, every client, every tool.
- First-class governed objects: Agents and Skills, not prompt snippets in someone's Notion page.
- Versioned + verified: diff, roll back, attest. Same Agent or Skill, every engagement.
- Auto-injected: clients discover and load them; no manual prompt hacks.
- Conditional load: attach only when the request semantically needs them.
- Fork + lineage: branch an Agent for a client engagement; trace every version's origin.
- Builder + Optimizer: meta-tools for authoring and tuning Agents and Skills.
One memory across every tool.
Entities, decisions, and context accumulate in airlock, not trapped in any single vendor. What you tell ChatGPT is what Claude Code sees tomorrow.
Queryable as structure.
airlock's memory is a DuckDB-backed entity graph of people, projects, decisions, accounts, and the relationships between them. Searchable, exportable, governed. Distinct from conversation-log memories that sit at the chat layer.
- Cross-vendor: one memory across Claude, ChatGPT, Cursor, Copilot.
- Searchable + governed: every entity attributable, every update logged.
The org's institutional brain.
Memory survives tool churn and team rotation. New hires inherit context; client engagements carry forward without re-briefing. 1.243 entities tracked across pilot tenants today.
Every action logged. Every action attributable.
Procurement-grade record of what every agent did, not just what was said. airlock sits on the action surface across vendors, so the audit log is cross-vendor by construction. The evidence trail an auditor signs off on, generated as you operate.
- Action audit: every tool call, approval, and Skill invocation logged with
agentInvocationIdcorrelation. - Matched-rule logging: each action stamped with the policy rule that allowed (or denied) it.
- A2A governance: agent-to-agent calls audited across heterogeneous runtimes. The only architecture that can.
- EU AI Act evidence pack: cross-vendor, article-mapped, exportable. EU-hosted on AWS Frankfurt; GDPR-aligned; ISO 27001 in progress.
- Project hierarchy: per-engagement audit boundaries with org-level inheritance, for agencies and consultancies running client work.
- SIEM-exportable and queryable from day one.
Governance as a route, not a roadblock.
AI adoption stalls on approval queues. airlock pre-approves safe actions, queues destructive ones, and stamps every event with the rule that allowed (or denied) it, so the audit log answers "who, what, and which policy" without a forensic exercise.
- Pre-approval policy engine: tag tools Read-only or Destructive, set rules once.
- Real-time approvals: route to Slack, Teams, or the Control Room. No tickets, no week-long queues.
- Matched-rule audit: every action stamped with the policy rule that fired. The procurement-grade trail auditors actually accept.
- Anomaly detection + kill switch per MCP server.
Anything not in the catalog? Upload an OpenAPI spec.
airlock generates the MCP server from your spec, OAuth-aware, governed by your policy. No engineering required. An Ops Manager or AI Practice Lead can stand up a new integration before lunch.
From spec to governed tool.
Drop in an OpenAPI YAML/JSON. airlock generates the MCP server, surfaces the tools, routes auth through the vault. Inherits per-tool policy, approvals, and audit from the moment it's live.
Built for the CPO without a dev team.
The original feedback from a design partner: "my CPO has no idea how to make a skill, let alone build governance." airlock solves it directly. Admin uploads the spec, the agent can now use the system.