Clarity at AI speed.

Imagine it.   Plan it.   Track it.   Ship it — with your agents.

The shared canvas where you and your agents turn intent into shipped work — never scattered, always in sync.

Watch one build
No login. The workspace link is the only key.
What it is

Crilo is the shared surface where a human and their AI agents plan, break down, track, and execute work together — at AI-native speed, organized by default, never scattered.

The loop · the heartbeat

Intent becomes a living visual plan.

01Imagine it
02Plan it
03Track it
04Collaborate on it
05Ship it
Loops — kept live by the agent

Intent becomes a living visual plan — mockups, technical diagrams, mermaid, charts — that you break down, track, work, reference, and adjust. The agent keeps it live, so it never goes stale.

Principles

How Crilo behaves.

01
Shared memory for your parallel agents.
The canvas is to your many sessions what main is to parallel git worktrees: read before acting, write as you go.
02
The unit is a Plan, not a board.
Every crilo belongs to a plan. No orphans, no scatter — enforced by the agent’s behavior and the model.
03
Collaboration in the loop, not autonomous bookkeeping.
For work humans need to stay aligned on, not a ticket/incident system of record. An agent can stand up a war-room crilo for an outage — collaboration, not autonomous tracking.
The bet · the moat
Crilo wins on craft: the best design, ideation, visualization, and organizational prompt engineering. The canvas is the substrate; the prompt layer — the MCP instructions + design guide — is the product.
Non-goals

What Crilo is not.

Not
A ticketing / incident system of record.
Not
A substrate for autonomous bot loops.
Not
“Another canvas you organize by hand” — the agent keeps it live.
Get started

Hand your agent the link.

Create a workspace

Mints a fresh, unguessable workspace and opens it. You’ll get an MCP snippet to connect your agent.

Open an existing workspace

Anyone with the id has full access — it’s the only credential. Bookmark your workspace to return.
$ claude mcp add --transport http crilo https://mcp.crilo.app/mcp
100%