Cortex

The description states that Cortex is a shared canvas tool that sits between teams and coding agents. It allows visual feedback on agent-built UI through annotations (dragging elements, redlining, sticky notes), which are compiled into precise specs and applied to real code by the agent.

• The agent pushes live screen captures with element maps.
• Teammates annotate visually using a desktop app or browser interface.
• Annotations are compiled into specs that the agent applies to code.
• The system uses peer-to-peer connections for collaboration without hosted servers.
• It includes a local Node process, MCP protocol integration, and a React-based UI with Fabric.js canvas.

Evidence Self-reported by authors. No independent verification.

The authors claim that working with coding agents has an "asymmetry" — the agent can generate UI quickly but feedback is limited to prose, which they describe as inefficient and unlike how designers work with humans.

They position Cortex as filling a missing gap in the agent workflow by enabling visual feedback loops. They also state that traditional chat-based feedback breaks when more than one person is involved, and that Cortex enables team sessions where everyone marks up UI on a shared canvas.

Evidence Self-reported claims about user pain points and product positioning.

The description implies the target customer includes:

• Designers
• Product managers (PMs)
• Developers working with coding agents
• Teams collaborating on UI built by AI agents

It suggests that feedback loops break when multiple people are involved in agent-assisted UI development, and that Cortex addresses this.

Evidence Self-reported. No explicit segmentation or customer data provided.

There is no evidence of pricing, monetization strategy, or business model in the description.

Evidence Not evidenced.

The system is described as:

• A single local Node process
• Uses MCP protocol over stdio to communicate with agents
• Serves UI via HTTP and WebSocket
• Built with React 19, Fabric.js, Fastify, Playwright, SQLite, Hyperswarm, Automerge, etc.
• Features a "latch" mechanism for synchronous feedback
• Employs peer-to-peer collaboration using Ed25519 keys, encrypted transfers, and CRDTs
• Screenshots are content-addressed and verified before display

Evidence Self-reported technical architecture and implementation details.

There is no evidence of traction, revenue, or adoption beyond the fact that it was submitted to a hackathon. The team size is listed as 4 members, but there is no mention of customers, users, or product usage metrics.

Evidence Not evidenced.

The description does not provide any information about competitors or market positioning relative to other tools in the space.

Evidence Not evidenced.

• The entire description is self-reported and unverified.
• No evidence of revenue, customers, or product usage.
• The tool appears to be a hackathon submission with no indication of commercial viability or market traction.
• Technical complexity (e.g., P2P, CRDTs) may not translate into real-world usability without further validation.
• The system relies heavily on agent behavior and protocol adherence — if agents don’t follow expected patterns, the UX breaks.

Inference If this is a prototype or proof-of-concept, it may not yet be ready for commercial use or investment.

1. What specific coding agents does Cortex integrate with?
2. Has the system been tested in real-world workflows beyond the hackathon?
3. Are there any early adopters or pilot users?
4. How does Cortex handle edge cases like DOM refactoring between agent versions?
5. What is the current roadmap for monetization or product development?
6. Can you demonstrate actual usage of the tool with a live agent session?