Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.cyberun.cloud/llms.txt

Use this file to discover all available pages before exploring further.

Cyberun exposes a Model Context Protocol (MCP) endpoint so external AI clients — Claude Code, Cursor, Windsurf, Codex CLI, Claude Desktop, VS Code 1.101+ — can connect to your team and call platform operations as MCP tools. This page is the cloud-side “what do I do as a user” walk-through. The complete reference — per-client config snippets, the full tool catalog, the error envelope — lives in MCP → Cloud MCP.

​
What you need

  1. An integration credential (sk-...) for the team you want the AI client to drive. Create one in Access → Integration — see Generate an API key.
  2. An MCP-aware AI client. Claude Code, Claude Desktop, Cursor, Windsurf, Codex CLI, and VS Code 1.101+ all support the Streamable HTTP transport the Cyberun MCP endpoint uses.
If the Access → Integration section doesn’t show MCP as an option for your team, MCP-based integration isn’t enabled on this deployment — contact your team admin.

​
The endpoint

https://core.cyberun.cloud/api/v1/mcp
Authenticate with the sk- key as a Bearer token. The team is bound to the credential; no other headers are required.

​
Configure your AI client

The exact config shape differs per client (config-file path, top-level key name, transport hint). The same URL and same Bearer header work for every one. Pick your client and copy the snippet: → Cloud MCP → Connect has the verified-current snippet for each of: Claude Code (CLI), Claude Desktop / Code (config file), Cursor, VS Code 1.101+, Codex CLI, and Windsurf (via mcp-remote shim). Restart the client after editing config. Cyberun tools appear in the tool picker.

​
What the client can do

Once connected, the client lists 15 tools that cover read-and-write operations on your team:
  • List and describe workflows.
  • List the team’s agents and connected container services.
  • Submit a task against a workflow.
  • Stream task progress and fetch result artifacts.
  • Cancel a running task.
Tools that mutate team-wide resources (issuing credentials, adding members) are not exposed over MCP — those require dashboard or direct API access. The full tool reference, inputs and outputs, lives at MCP → Cloud MCP → Tools.

​
Help the agent use it well

If you drive Cyberun from an AI agent, install the Cyberun skill so the agent knows the API and MCP layout without you re-explaining: 
npx skills add cyberun-cloud/skills
The skill teaches the agent which tools exist, how to authenticate, when to stream vs poll, and the common mistakes to avoid.

​
Revoke an MCP client’s access

When a client should no longer drive your team — laptop lost, key suspected leaked, agreement ended — revoke the credential:
  1. In Cloud, open Access → Integration.
  2. Find the credential by its name (full key values aren’t shown after creation).
  3. Click Revoke.
The MCP connection returns 401 on the next call. The client should prompt the user for a new credential.

​
Troubleshooting

Client doesn’t list Cyberun tools. MCP isn’t enabled for this deployment — the Access page won’t show an MCP entry in that case. Ask your team admin. 401 immediately. The sk- key is wrong, revoked, or expired. Create a fresh one in Access → Integration. Tool list is empty. The credential is valid but the team has no workflows yet. Add one in Cloud first. Client hangs on long tasks. Streamable HTTP delivers progress events incrementally. If your client doesn’t render them, the task is still running — check it at app.cyberun.cloud/tasks. For more, see MCP → Cloud MCP.