Claude Code

Put the brief in One Horizon first. The initiative, bug, or todo should give Claude Code the goal, context, links, comments, and acceptance notes before it starts.

After that, choose how you want to work. Stay in Claude Code with the MCP connector, or let a local Claude Code agent run the task on your machine from Desktop, the CLI, or Workflows. The work item carries the prompt detail.

Local agent setup is separate from the MCP connector. It uses your authenticated claude CLI and does not require you to add the One Horizon MCP server manually.

Choose how to use Claude Code

Start with the work item

Claude Code works best when the initiative, bug, or todo already explains the change. Add the outcome, important files or links, constraints, and what counts as done. That work item becomes the prompt source for both MCP chat and local agent runs.

For larger implementation work, keep the brief close to spec-driven development: specific enough for Claude Code to act, short enough for a person to review.

Connect Claude Code through MCP

MCP is the interactive path. Configure it when you want to ask Claude Code about One Horizon work while you are already coding.

The Claude Code integration uses the One Horizon remote MCP server over HTTP with OAuth.

Import from Claude Desktop

If you already configured the MCP server in Claude Desktop, import it.

claude mcp add-from-claude-desktop --scope user

Configure manually

claude mcp add --transport http onehorizon https://mcp.onehorizon.ai/mcp

Manage servers

List, inspect, remove, or check status.

claude mcp listclaude mcp get onehorizonclaude mcp remove onehorizon/mcp

Authenticate

When you run /mcp in Claude Code, a browser window opens so you can sign in and authorize the connection. After granting access, return to Claude Code.

/mcp

Use Clear authentication in the /mcp menu to revoke.

Project level

Share a team configuration by committing a .mcp.json at the repository root. Our MCP server uses OAuth so there is no need for environment variables.

{  "mcpServers": {    "onehorizon": {      "type": "http",      "url": "https://mcp.onehorizon.ai/mcp"    }  }}

Run a Claude Code local agent

A local Claude Code agent is for tasks you want Claude Code to work through without a live chat. After setup, open an initiative, bug, or todo in One Horizon, choose Send to, and pick your Claude Code agent. The run stays tied to that work item.

When a local Claude Code agent run reaches complete or error status, the session stores a Claude session ID. Open in Claude Code appears in the run menu and resumes that session in the task workspace folder. The action is available only when the session status is complete or error so it does not start a separate interactive session while work is still active.

Setup, agent commands, and the execution model are described in Local Agents and Agent Sessions.

Ask Claude Code through MCP

Short prompts are enough when the work item already has the detail:

Try these prompts:

"show my assigned issues and todos for today"

Claude Code pulls your in-flight issues from tools like Linear or Jira and summarizes them inline.

"use my assigned work details for the gateway API"

Claude Code brings those details into context so the AI names things correctly and scopes changes.

"capture this as completed work"

Claude Code summarizes what it generated and adds a completed task.

Security and permissions

The connector acts on your behalf. Data access follows your roles, workspace, team membership, and scopes.

Troubleshooting