Connecting External AI Tools via MCP

Septra supports the Model Context Protocol (MCP), an open standard designed to link AI assistants, development tools, and IDEs directly to databases and applications. By connecting your local AI clients to Septra's workspace MCP endpoint, you can search your Knowledge Base and invoke your agents directly from your terminal or coding assistant.

---

What is the Septra MCP Server?

Septra exposes a secure Streamable-HTTP MCP endpoint at POST /mcp on the main REST API.

This endpoint sits behind our workspace authorizer. When an external AI client connects, it authenticates via secure workspace OAuth protocols. Once validated, the client gains access to the following workspace tools:

• search_knowledge_base: Queries your workspace's semantic memory and returns citation passages.
• list_agents: Returns the IDs, roles, and descriptions of all AI agents deployed in your workspace.
• ask_agent: Sends messages directly to your agents and retrieves their responses.

---

1. Connecting to Claude Code (CLI)

The Claude Code command-line interface natively supports HTTP MCP transport and PKCE OAuth flows.

To link your terminal assistant to your Septra workspace:

1. In your terminal, run the following command (substituting the correct API endpoint for your environment):

`bash

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

`

1. The Claude CLI will detect the OAuth verification requirements and trigger a secure login page in your default browser.
2. Sign in using your workspace email and password.
3. Once authenticated, the browser window will authorize the CLI client, and the connection dot will turn green in your terminal console.
4. You can now use Septra tools in your conversations:
   • *"Claude, check Septra's knowledge base for the product voice guide."*
   • *"Claude, ask our CMO agent Nova to review my current codebase directory."*

---

2. Connecting to Claude Desktop

Claude Desktop expects standard stdio input/output configurations. To bridge Claude Desktop to Septra's HTTPS MCP endpoint, you can use the mcp-remote proxy bridge.

1. Install the bridge proxy on your local machine:

`bash

npm install -g mcp-remote

`

1. Open your Claude Desktop configuration file:
   • macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
   • Windows: %APPDATA%\Claude\claude_desktop_config.json
2. Add the server parameters to the mcpServers block:

`json

{

"mcpServers": {

"septra": {

"command": "npx",

"args": [

"mcp-remote@latest",

"--sse",

"https://api.septra.ai/mcp"

]

}

}

}

`

1. Completely close and restart the Claude Desktop application to reload the settings.

---

3. Connecting to Cursor IDE

Cursor supports direct SSE and HTTP configurations:

1. Open Cursor Settings > Features > MCP.
2. Click Add New MCP Server.
3. Configure the parameters:
   • Name: septra
   • Type: SSE
   • URL: https://api.septra.ai/mcp
4. Click Save and wait for the status indicator to turn green.
5. In your chat sidebar, you can now toggle the @septra tool registry to ground Cursor's programming logic in your workspace's actual documentation.