[ REDPAPER ]  September 2026

1Claw Robotic RedpaperThe access layer for robots

Ten guardrails we ship today, and ten we are building next, for robots that need to touch human systems, walk into human places, and wait on human approval.

September 2026 16 pages 20 guardrails

A Digit unit writing an inventory count back into a warehouse system. A Spot unit approaching a switchgear room. A hospital robot at a locked drug cabinet. None of these are motion problems — every one is an access request, made by a machine that was not a customer of your identity systems five years ago.

Read the Redpaper

Twenty guardrails with the actual call shapes — Cedar policies, binding configs, approval payloads. We email you a download link.

By submitting, you agree to 1Claw's Privacy Policy. We never sell your data.

01–10 · Ships today

Running in the product now, with the shapes of the calls involved.

  1. 01Never hand a robot a raw credential
  2. 02Broker access to WMS, MES, EHR and CMMS
  3. 03Inspect what a robot reads before a model sees it
  4. 04Park high-risk actions for human approval
  5. 05Scope every unit to its own short-lived token
  6. 06Govern the fleet without silent drift
  7. 07Detect anomalies and revoke on the spot
  8. 08Cap what a robot can spend
  9. 09Make a retried action execute exactly once
  10. 10Govern human operator and remote takeover access

11–20 · Roadmap

A build plan we want to shape with the first robotics teams running this in production — not a roadmap announced after the fact.

  1. 11Approval classes built for physical risk
  2. 12Tie approvals to a real work order
  3. 13Vend physical credentials like API keys
  4. 14Signed operating envelope tokens
  5. 15Bind identity to the robot's hardware
  6. 16Bounded autonomy when the network drops
  7. 17Cohort quarantine on a shared signal
  8. 18Injection defence for cameras, not just text
  9. 19Real governance for teleoperation sessions
  10. 20Multi-party governance on one site

Why a redpaper

A whitepaper usually describes a problem someone already solved. We have not finished solving this one. Ten of the twenty items are shipping today; ten are a build plan. Red is also our colour, and pretending this were a finished story felt worse than just saying so.

What we will not build: motion control, torque limiting and emergency stop stay inside the robot's certified safety controller, permanently. We are not applying for that job, and a network-dependent API does not belong near it. We govern who gets a credential, and under what approval.

Every permission traces back to a human who granted it. That does not change because the thing asking for permission now has arms.

If one of the ten roadmap items matters more to your fleet than we have guessed, tell us and we will build that one first.