Human-in-the-loop for Claude Code
Claude Code runs long, autonomous sessions in your terminal, which is exactly when it hits product and architecture calls it shouldn't make alone. Add Orako as an MCP server and Claude Code can pause, ask the teammate who owns the domain, and continue with a real answer instead of a confident guess. The reply lands back in the session, and the exchange is saved so nobody answers the same question twice.
Connect Orako to Claude Code
- 01
Run
claude mcp add --transport http orako https://app.orako.io/mcp, or merge the snippet below into .mcp.json. - 02
Approve the connection with
/mcpinside Claude Code. Claude Code discovers theaskandsearch_historytools at the same time. No prompt engineering. - 03
When it hits a human-only decision, it asks through Orako, waits for a reply, and resumes.
You approve access inside Claude Code itself — no token to copy, nothing written to a local credentials file.
{
"mcpServers": {
"orako": { "type": "http", "url": "https://app.orako.io/mcp" }
}
}User scope lives in ~/.claude.json under the same mcpServers key. Approve access with /mcp inside Claude Code — the agent's guidance ships with the server itself, no skill file to install.
New to MCP in Claude Code? See Anthropic’s docs.
What Claude Code can ask
Let Claude Code ask a human before it guesses.
Schema decisions mid-refactor
Claude Code is halfway through a migration and needs to know whether soft deletes are a product requirement. It asks the data owner instead of picking a side, and the refactor lands right the first time.
Product rules before edge cases
An invite flow has an ambiguous seat-limit rule. Instead of encoding a guess in code and tests, the agent asks whoever owns billing and implements the confirmed behavior.
API contracts before shipping
Renaming a public endpoint field is cheap now and expensive later. Claude Code gets a sign-off from the API owner while it keeps working on the rest of the change.
Claude Code + Orako, answered
Does Claude Code support MCP natively?
Yes. Claude Code has first-class MCP support: project-scope servers in .mcp.json at the repo root, user-scope servers in ~/.claude.json. Orako connects as a remote HTTP server, so there's no binary to install; approve access once with /mcp and its tools appear in the session.
What does Claude Code do while it waits for an answer?
The ask call is a normal tool call: Claude Code keeps working on other parts of the task if it can, or waits for the reply. When a teammate answers in Slack, Discord, or the dashboard, the answer returns as the tool result and the session continues.
Do my teammates need this agent installed to answer?
No. Only the developer running the agent connects Orako. The people answering receive each question in Slack, Discord, or the Orako dashboard and reply in the thread. They never touch an editor, a terminal, or an MCP config.
What happens to the answers?
Every resolved conversation becomes searchable team history with its authors, date, tags and full context. Agents call search_history before asking, so a question your team already answered can be reused instead of pinging anyone again.
Put a human in the loop for Claude Code
Give Claude Code the one thing it can’t generate: the answer only your team knows.