Castellan protects your personal context across AI agents.

Keep your experience. Level up with every agent.

An invitation and a passkey are required. See how to get started. The overview PDF is public and shareable.

Every agent starts at zero. Castellan captures your personal context layer securely.

Start in your browser at beta.castellan.us. Save the knowledge and rules you approve, use them with your personal Squot, and choose what each connected agent can read. Your online workspace stays in encrypted account storage.

Illustrated example: one preference shared with Cursor after owner approval, with a read recorded in Watchtower
Illustrated example using fictional context. You choose which agents may read it.

YOUR PROGRESS CARRIES FORWARD

Your experience is the upgrade.

Your next agent should inherit your progress. Keep the preferences, corrections and working methods you approve, then share the relevant parts with each new agent.

  1. 01 · REMEMBER

    Build your base

    Approve a useful memory once. Let your connected agents draw on it with the access you choose.

  2. 02 · REFINE

    Sharpen your playbook

    Turn a correction into a standing rule. Keep its source and history, and refine it as your needs change.

  3. 03 · ADVANCE

    Start further ahead

    Carry decisions, completed work and next steps into your next agent conversation with a reviewed handoff.

The upgrade stays with you. Your context and playbook grow. You choose what each agent inherits.

Explore how your knowledge compounds

A castle for your context

An approved agent connection reaches the context broker with scoped access and an expiry. The workspace keeps approved records in encrypted account storage. Your Squot uses selected context. Data controls provide explicit export and removal. Watchtower shows context access and agent-reported actions.

Two layers in one ledger

Who you are

  • identity
  • people
  • goals
  • habits
  • preferences
  • routines

Facts you chose to write down or confirm, each with its date and source.

How you want agents to work

  • standing rules
  • style
  • autonomy limits
  • privacy lines
  • procedures

Your approved rules accompany context reads. They guide an agent's behavior; Castellan enforces access to your context, while the agent controls its work in other services.

Build your context layer. Choose access. Give your next agent a head start.

Three steps: build your context layer, choose what each agent can access, and carry approved knowledge into your next conversation.

Castellan updates

Stay connected.

Receive Castellan news by email. Unsubscribe at any time. Updates do not create a beta account.

Your address is stored in Castellan's mailing list, separately from beta accounts. Privacy

Your five rooms

Overview shows your next step. Records keeps approved context. Agents manages connections, with the Gatehouse for approving or refusing access. Watchtower shows the evidence. Settings holds your account and security choices.

Explore the rooms and the approval flow