How you customise it

1. Open Resources to find your instructions, skills, and integrations together. Filter by where they come from: your own, shared with you, from Tracelight, or set for a group or project.

   

2. Each tab explains what it holds, with Upload and Create to add your own. On Skills, style guides for General, Excel, and PowerPoint sit at the top, with other skills below.

   

Type “/” in the chat to open the Skills menu. Use or build skills to automate repetitive tasks - to build one, run /create-skill. To use a skill, type “/” and select it. Share skills from the Resources tab.

Instructions are always-on rules the agent must follow, use them for behaviour you want applied to every task.

1. Access instructions via the Resources tab to view and create them.

   

2. Instructions are Markdown files you can edit, and apply to specific folders.

   

3. To toggle instructions on and off, click the “+” in the chat to see which instructions are active.

   

1. Open the Skills tab in Resources. Style guides sit at the top, grouped by General, Excel, and PowerPoint. You can toggle on/off which ones you want active.

   

2. Click “Create style guide” and upload a workbook or deck. Tracelight extracts its formatting conventions - fonts, colours, number formats, chart conventions - into a reusable style guide applied to every output.

   

3. You can also toggle style guides on and off from the chat: click the “+” button and open Style guides to see what is active for this conversation.

   

Open Integrations to connect your data sources: Microsoft 365 for SharePoint & OneDrive, Microsoft Work IQ, S&P Capital IQ, FactSet, PitchBook, and more. Once connected, @mention a source in the chat and Tracelight pulls the data straight into your model, cited. Tracelight also connects to almost any other tool through MCP; see Connect other tools via MCP.

MCP (Model Context Protocol) is the open standard most tools use to expose their data and actions to AI agents. Each MCP server you connect adds its tools to Tracelight’s agent: it can search your Notion pages, read a Slack thread, pull figures from FactSet, or create a Linear issue, using your own account on that service.

Add a server from the catalog

1. Open Integrations and find the tool under Other integrations: Microsoft Work IQ, S&P Capital IQ, FactSet, PitchBook, LSEG, Morningstar, Notion, Slack, Linear, Granola, and more.

2. Click “Add” on its card. If the server needs you to sign in, click “Connect” next to it and sign in to that service.

3. The green dot and tool count show the server is ready. Expand the count to see its tools, and use the switch to turn the server off without removing it.

Add any other MCP server

Click “Custom MCP server” (or “Edit configuration”) to open your MCP configuration, add an entry with the server’s address, and save:

{
  "mcpServers": {
    "Linear": {
      "type": "http",
      "url": "https://mcp.linear.app/mcp",
      "allowedTools": ["list_issues", "create_issue"]
    }
  }
}

allowedTools is optional: leave it out to give the agent every tool the server offers. Servers that sign in with OAuth show a “Connect” button once saved. For a server that takes a static token, set an Authorization header under headers instead. A server that doesn’t support dynamic client registration needs an oauth.clientId and oauth.clientSecret; register the redirect URL shown in the configuration dialog with the provider first.

Use it in a chat

Every server you have turned on is available to the agent, on the Platform and in the add-ins. Ask for what you need in plain language and the agent calls the server’s tools when they help. To turn servers on or off without leaving the conversation, click the “+” button in the chat and open Integrations.

Tracelight runs an MCP server, open on every plan, so you can use it from Claude, ChatGPT, Cursor, and any other MCP client. From your AI tool, you can hand tasks to the Tracelight agent, browse and upload project files, and collect what the agent produced.

Every client connects to the same address over streamable HTTP. The first time you connect, a browser window opens for you to sign in with your Tracelight account; there is no API key to copy.

https://app.tracelight.ai/api/mcp

Claude (web and desktop)

1. In Claude, open Settings → Connectors and click “Add custom connector”.

2. Name it Tracelight, paste the server address above, and click “Add”. Then click “Connect” and sign in to Tracelight.

Claude Code

1. Run this in your terminal:

   claude mcp add --transport http tracelight https://app.tracelight.ai/api/mcp

2. Start Claude Code, type /mcp, choose tracelight, and select “Authenticate” to sign in.

ChatGPT

1. In ChatGPT, open Settings → Apps & Connectors → Advanced settings and turn on Developer mode.

2. Back in Apps & Connectors, click “Create”, name it Tracelight, paste the server address above, choose OAuth authentication, and create it. ChatGPT then asks you to sign in to Tracelight.

3. In a new chat, pick Tracelight from the tools menu under the composer.

Codex

Run these commands. The second opens a browser window to sign in to Tracelight.

codex mcp add tracelight --url https://app.tracelight.ai/api/mcp
codex mcp login tracelight

Cursor

1. Open Cursor Settings → Tools & MCP and click “New MCP Server”. This opens ~/.cursor/mcp.json; add this entry and save:

   {
     "mcpServers": {
       "tracelight": {
         "url": "https://app.tracelight.ai/api/mcp"
       }
     }
   }

2. Back in Tools & MCP, click “Connect” next to tracelight and sign in.

VS Code (GitHub Copilot)

1. Open the Command Palette, run “MCP: Add Server”, choose HTTP, paste the server address above, and name it tracelight. Alternatively, add this entry to .vscode/mcp.json in your workspace, or to your user mcp.json via “MCP: Open User Configuration”:

   {
     "servers": {
       "tracelight": {
         "type": "http",
         "url": "https://app.tracelight.ai/api/mcp"
       }
     }
   }

2. VS Code asks you to sign in to Tracelight when the server first starts. Use Copilot Chat in Agent mode to call its tools.

Other MCP clients

Any client that supports remote MCP servers over streamable HTTP with OAuth can connect: give it the server address above and sign in when prompted. If a client only supports local servers, bridge it with mcp-remote by running the server as:

npx -y mcp-remote https://app.tracelight.ai/api/mcp