Connect ConvOps to Claude Code

What do you get in Claude Code once ConvOps is connected?

Claude Code keeps doing the work in your repository. ConvOps holds the process around it, so the session follows the same steps every time instead of whatever the prompt implies today.

1. Tasks that carry a workflow. Claude asks the engine for the current step, does it, and advances.
2. Gates the engine evaluates. An unmet condition is returned to Claude and the run does not move.
3. Approval steps that wait for you, in the same terminal, before anything ships.
4. Memory and routes shared by every session and every teammate, so context survives /clear and a new day.
5. An audit row for each change, attributed to the agent and to you.

ConvOps is the operations layer for AI agents: an MCP server that holds your team's process as workflows, with approval gates, a shared memory and an audit trail. It runs no AI models; Claude Code does the work. The steps, gates and decisions come from the ConvOps workflow engine, and the approvals and audit trail from governance.

What do you need to connect Claude Code to ConvOps?

A Claude Code setup that supports remote MCP servers, a ConvOps account, and the server URL. Plans start free; see pricing for limits.

How do I add the ConvOps MCP server to Claude Code?

Three commands. The first writes the server into your Claude Code config; the second signs you in; the third confirms it.

1. 1

   Add the server

   Run this in any terminal. --scope user makes ConvOps available in every project; leave it out to add it to the current project only (the default local scope).

   claude mcp add --transport http --scope user convops https://mcp.convops.app/

2. 2

   Sign in from inside Claude Code

   Start claude, type /mcp, pick convops and choose to authenticate. Your browser opens the ConvOps sign-in; approve, and the token is stored and refreshed by Claude Code.

   /mcp

3. 3

   Check the connection

   The server should show as connected. "Needs authentication" means step 2 has not finished.

   claude mcp list
   # convops: https://mcp.convops.app/ (HTTP) - ✔ Connected

4. 4

   Start a task

   Ask in plain words, for example "Create a bug task for the checkout timeout and start its workflow." Claude calls ConvOps, the workflow attaches, and the first step arrives.

claude mcp add --transport http convops https://mcp.convops.app/ \
  --header "Authorization: Bearer $CONVOPS_API_KEY"

Official reference: Claude Code: connect to MCP servers, Claude Code: memory and CLAUDE.md. We checked these steps against those docs on Oct 4, 2026. Anthropic updates its interface often; when a menu name differs, the vendor docs are current.

What does a governed task look like in Claude Code?

A bug fix, walked one step at a time. Claude never sees the whole process at once: it gets the current step, does it, and asks to advance. At the approval step the run waits for you.

Each tool row is Claude Code calling ConvOps. The engine hands over one step, checks the gate when the agent asks to advance, and holds the run at an approval step until a person answers. The same task can be picked up later, by you or by a teammate in another client, at the step where it stopped.

What should Claude Code users know about CLAUDE.md, scopes and unattended runs?

Keep CLAUDE.md short. ConvOps sends its own usage instructions on connect and each step arrives with its own, so one line such as "Use ConvOps for task work and follow the workflow step by step" is enough. See CLAUDE.md too long for what to move, and make Claude Code follow a process for prompts versus hooks versus gated workflows.

Scope decides who sees the server. --scope project writes it to .mcp.json in the repository, so the whole team gets ConvOps on clone; each person still signs in with their own account and Claude Code asks for approval before using a project server the first time.

Claude Code loads MCP tools through tool search by default, so the ConvOps tool list does not crowd the context until a tool is needed.

Attended and unattended are the same workflow. Attended means you run it in this terminal and answer the gates. Unattended means an executor dispatches the task to an isolated run, one pod per run on your own Kubernetes, which runs Claude Code or OpenCode today; the gates still stop it. See run Claude Code unattended on Kubernetes.

Dynamic workflows and ConvOps do different jobs. A Claude Code dynamic workflow fans out many subagents inside one run; a ConvOps workflow holds the stages around it and the sign-off between them. The Dynamic Workflows comparison shows how to use both.

Why is ConvOps not showing up in Claude Code?

Almost always one of two things: the sign-in has not finished, or the server is off for this session. Start with the first row.

What else do people ask about ConvOps in Claude Code?

Short answers, checked against the Anthropic documentation linked above.