The MCP security gateway
Hodor is the security gateway for AI agent governance — between your harnesses and your applications, making sure each agent has the right tools for the right job. Nothing more.
Every tool call your agents make, brokered and logged.
| Time | Status | Operation | Latency |
|---|---|---|---|
14:32:18Jun 21 | 200 |  GitHub|Claude Code · Platform create_issue alex@northwind.io · Engineering | 240ms |
14:32:11Jun 21 | 200 |  Zendesk|Support Copilot get_ticket PII redacted priya@northwind.io · Support | 180ms |
14:32:04Jun 21 | 403 |  Gmail|Outbound SDR send_bulk_email policy blocked lea@northwind.io · Revenue | 12ms |
14:31:59Jun 21 | 200 |  BigQuery|Data Analyst run_query daniel@northwind.io · Data Platform | 910ms |
14:31:52Jun 21 | 200 |  Linear|Cursor · Frontend list_issues marco@northwind.io · Engineering | 156ms |
14:31:47Jun 21 | 200 |  Stripe|Support Copilot get_customer PII redacted priya@northwind.io · Support | 320ms |
14:31:40Jun 21 | 403 | GitHub|Claude Code · Platform delete_repo policy blocked alex@northwind.io · Engineering | 8ms |
Backed by
Integrations
What is Hodor
Every existing path — OAuth, bearer tokens, app-level scopes — was designed for humans clicking buttons, not agents running thousands of calls a minute, so you over-permission by default. Hodor sits in front with one connector that handles identity, permissions, policies and observability: each agent gets exactly the access it needs — for fewer tokens, a smaller blast radius, and the record to prove it.
WithoutThe agent inherits your OAuth grant — it can do anything you can, with no sub-identity to scope or revoke.
WithA non-human identity per agent, in your registry and linkable to your IDP, with short-lived, scoped keys.
WithoutApp scopes were built for a human clicking buttons — far too broad for an agent making thousands of calls.
WithTool scope and payload scope per agent. It can't discover or call what isn't allowed.
WithoutNowhere to express a rule on a payload or across apps — access is all-or-nothing.
WithPolicy-as-code on the wire — PII redaction, argument constraints, and cross-app rules, enforced every call.
WithoutEvery tool the agent can see is a tool it can call by mistake — one poisoned input or over-eager plan and it fires.
WithAllowed behavior mapped to expected behavior. Most misuse simply can't be expressed.
WithoutEvery tool definition ships in the prompt on every turn — 90 tools instead of 12 burns thousands of extra tokens per loop.
WithOnly the tools the job needs — up to 80% less in tool-token spending across the run.
WithoutTool schemas eat the context window before the agent does anything useful.
WithScoped tools free the context for the task — up to 80% less context spent on tooling.
WithoutKeys reused across agents and tasks — when something breaks, you can't tell who did what.
WithA replayable audit trail for every call, tied to the identity that made it.
Talk to us
Bring an agent you're about to deploy. In fifteen minutes we'll show you what it can actually reach today — and what it can reach once Hodor is in front of it.
