Skip to content

One capability

AI agents you can verify, in conversations we cannot read.

Invite an agent with its own keys into a Space. Check its identity and who operates it before sharing client information.

Three people working at a long table with empty message panels above them, and a faceted solid on a stand at the fourth place, drawn as one continuous line

A seat at the table

An invited AI agent holds its own keys and decrypts its own copy of a conversation. The relay still carries ciphertext. The agent’s operator and any connected model provider can process the content it receives.

Who is in the room

Illustrative membership for a mortgage application: four people and one drafting agent. Roles and permissions are configured for the deployment.

The borrower
HoldsOwn keys · verified
MaySend · upload · sign
The broker
HoldsOwn keys · verified
MaySend · upload
The solicitor
HoldsOwn keys · verified
MaySend · upload · sign
A compliance officer
HoldsOwn keys · verified
MaySend · approve
A drafting agentAgent
HoldsOwn keys · verified
MaySend · draft · ask for approval

Every row holds the same kind of keys, and every row is visible to everyone else in the room. The last one is labelled, because a participant nobody can see is not one anybody agreed to.

Checking it is the same job

Check the agent’s identity, the organisation operating it and its Space membership. Cryptographic verification helps detect changed keys; it does not establish that an agent or operator is trustworthy.

Two phones held side by side showing the same grid of verification digits, drawn as one continuous line
Two hands meeting at an approval card above a phone, drawn as one continuous line

An AI agent asks. A person decides.

For consequential actions, require an authorised person’s approval and check the exact terms before acting. Permissions and enforcement depend on the agent and deployment.

What it can read

An invited AI agent holds its own keys and decrypts its own copy of a conversation. The relay still carries ciphertext. The agent’s operator and any connected model provider can process the content it receives.

Only the rooms you put it in
An agent receives content through its Space membership and permissions. Review both before adding it, and review what it sends to external tools or model providers.
Review its activity
Agent tooling includes tool-call audit recording and argument redaction. Check the deployment’s audit configuration and retained records; redaction does not guarantee that every sensitive detail is removed.

Your agent. Your choice of model.

InTouch provides the identity, membership and encrypted channel. Your organisation chooses the agent and its model provider. MCP connects agent tooling; changing models does not require handing message decryption keys to the relay.

MCP · Ed25519 request signing · per-agent rate limits · agent accounts only

Taking it out again

Removing a member and advancing the Space’s group keys protects future messages under the updated state. It does not erase content, keys or copies the member already received.

How a Space works

Choose the agent. Check who runs it.

Choose who operates the agent, check its identity and review its permissions. The encrypted channel protects delivery; the agent and its connected providers process the content you make available to it.

How the encryption worksWhat a Space is