MCP

Connect Lace to Claude Code, Cursor, or Codex so your coding agent can read approved decisions and ship the fix.

Use Lace with your coding agent. When you log and approve a decision in the Decision Canvas, it becomes a scoped change request with evidence, spatial context, screenshots, and a suggested fix. The MCP server lets your coding agent pull those decisions, see exactly what needs to change, and write the code.

Get started

Install Lace MCP for your coding tool:

Cursor

Or install from the terminal:

npx lace-mcp install cursor

Restart Cursor, then open a new chat and ask about approved Lace decisions.

Claude Code

npx lace-mcp install claude

Start a new Claude Code session and say “List Lace decisions”.

Codex CLI

npx lace-mcp install codex

Start a new Codex session and say “List Lace decisions”.

The MCP server only works while the Lace desktop app is open and you’re signed in.

How it works

1. Log a decision

   In the Decision Canvas, capture a screenshot, select UI elements, and approve a decision. It persists with the screenshot, spatial context, and comments.

2. Your agent queries Lace

   It calls list_decisions, picks one, then calls apply_decision to get screenshots, targeting, and step-by-step instructions.

3. The agent ships the change

   It writes the code change and you review the diff.

Technical reference

MCP server

The Lace MCP server is registered under the name lace and exposes three tools:

list_decisions

Returns all approved product decisions as a summary list.

Parameters: none

Returns:

Field	Type	Description
id	string	Unique decision identifier
headline	string	Main decision headline
insightHeadline	string | null	The specific insight that was approved
elementEid	string | null	Element ID within the UI
appName	string | null	Application where the decision applies
approvedAt	string	ISO timestamp of approval

---

get_decision

Returns full details for a single decision — the card, element targeting, bounding box, and inline screenshots. Use this for browsing and inspecting decisions.

Parameters:

Parameter	Type	Description
id	string	Decision ID from list_decisions

Returns:

Field	Type	Description
id	string	Decision identifier
card	object	Decision card with headline, body, evidence, and insights
insightIndex	number	Which insight was approved (index into card.insights)
elementEid	string | null	Target element ID
bbox	array | null	Normalized bounding box [left, top, right, bottom] in 0-1 coordinates
screen	object	{appName, windowTitle, shotId}
approvedAt	string	ISO timestamp

Screenshots (raw and annotated) are returned as inline JPEG images when available.

---

apply_decision

Returns everything needed to implement a decision: screenshots, the decision card, element targeting, and step-by-step implementation instructions. Call this instead of get_decision when your agent is implementing a change.

Parameters:

Parameter	Type	Description
id	string	Decision ID from list_decisions

Returns:

• Annotated screenshot (JPEG, when available)
• Raw screenshot (JPEG)
• Implementation prompt with:
  • Decision card (headline, approved insight, suggested fix, evidence)
  • Target element (element ID, bounding box, screen info)
  • Step-by-step instructions to locate the element and implement the fix
  • Constraints on what not to modify

---

Config file locations

Tool	Config file
Claude Code	~/.claude.json
Cursor	~/.cursor/mcp.json
Codex CLI	~/.codex/config.toml

Troubleshooting

Agent can't connect to Lace

The Lace desktop app must be open and signed in. Open the app and retry.

No decisions returned

You need to approve nudges first. Open Lace, review the insight cards, and approve at least one.

Agent doesn't recognize the MCP server after install

Restart your coding tool fully (quit and reopen), then confirm the Lace server appears in your MCP settings.

Screenshots are missing from decision details

Screenshots are stored locally alongside the Lace app data. If they were deleted or the app was reinstalled, the decision will still return card and targeting data but without images.