Integration

Cursor Cloud Agents integration

Cursor Cloud Agents are AI coding agents that work on your repositories in Cursor-hosted VMs (or your own workers) and push their changes to a branch, optionally opening a pull request. Launch an agent from a prompt on one or more repositories, send it follow-ups or steer a running turn, read each run's status, final reply, pushed branches and pull requests, list its artifacts and get download links, read token usage, archive, unarchive, delete it, and list the models and GitHub repositories it can use. Manage the saved environments agents run in: create, read, update and delete them, read their change history, and start and read their builds. List the names (never the values) of Cloud Agents secrets. Left out: streaming a run (server-sent events), setting and deleting secrets, minting worker tokens, the self-hosted worker and pool routes, and the per-agent inline MCP servers and session environment variables (both carry credentials). Every agent run spends the user's Cursor usage at the chosen model's rate, and an agent pushes commits to the user's repositories. The Cloud Agents API is in public beta and needs a plan with Cloud Agents. Authentication is a Cursor API key (user or service account key from the Cursor dashboard, API Keys page) sent as a Bearer token.

Connect Cursor Cloud Agents How integrations work

Ask it in chat

Say what should happen.
The workflow builds itself.

When this happens, do that in Cursor Cloud Agents.Built from the 29 Cursor Cloud Agents tools below, on the connection you add once.

What you can do with Cursor Cloud Agents

Each of these is one operation on one endpoint, available both as a workflow step and as a tool an AI agent can call on its own. Nothing bundles several operations into one action, so what a run did stays readable afterwards.

Most endpoints run on the free plan. A few, mainly the ones that publish on your behalf, need a paid plan.

Environment, 5 operations

create_environmentCreate environmentPOSTCreates a saved environment, personal or shared with the team, with its repositories and its environment.json configuration (install and start commands).
delete_environmentDelete environmentDELETEPermanently deletes a saved environment. This cannot be undone.
get_environmentGet environmentGETReturns one saved environment with its repositories, its environment.json configuration (as a JSON-encoded string), the repository file it reads it from, if any, and its build settings.
list_environmentsList environmentsGETLists the saved environments (repositories plus setup) the API key can use, most recently updated first: personal and team ones. Filter by name text, owner, repository or user.
update_environmentUpdate environmentPATCHUpdates an environment; only the fields sent change. environmentJson replaces the whole configuration, repos replaces the whole list (send environmentJson with it), and buildSettings must be sent alone.

Agent, 4 operations

create_agentCreate agentPOSTLaunches a Cursor cloud agent that starts working at once. Each run spends the user's Cursor usage at the model's rate, and the agent pushes commits to a branch of the given repositories (or runs with none) and can open a pull request.
delete_agentDelete agentDELETEPermanently deletes an agent and its history. This cannot be undone; archive_agent is the reversible alternative.
get_agentGet agentGETReturns one agent: name, status (ACTIVE or ARCHIVED), environment, repositories, branch and pull request settings, custom subagents and latest run id. Run progress lives on runs (get_run).
list_agentsList agentsGETLists the cloud agents of the API key's user, newest first, with their status (ACTIVE or ARCHIVED), environment, web URL and latest run id. Filter by pull request URL or leave archived agents out.

Build, 3 operations

get_environment_buildGet environment buildGETReturns one build of an environment: status, trigger, source, draft flag, failure reason, requested branches, commits built and timestamps.
list_environment_buildsList environment buildsGETLists an environment's builds, newest first, with status, trigger, source, draft flag, failure reason and the commits built. Filter by status or creation time.
start_environment_buildStart environment buildPOSTStarts a build of an environment from each repository's default branch, or from given branches as a draft build that agents do not use until it is activated. Returns the build id.

Run, 3 operations

create_runCreate runPOSTSends a follow-up prompt to an agent; the run spends more Cursor usage and can push more commits. By default it starts a new run (409 while another run is active); with delivery set to steer it goes into the turn in progress instead.
get_runGet runGETReturns one run of an agent: status (CREATING, RUNNING or WAITING_FOR_BACKGROUND_WORK while active; FINISHED, ERROR, CANCELLED or EXPIRED once ended), duration, the agent's final reply and its pushed branches and pull requests.
list_runsList runsGETLists the runs of an agent, newest first, each with its status, duration, final reply and the agent's pushed branches and pull requests.

Secret, 3 operations

list_environment_secretsList environment secretsGETLists an environment's Cloud Agents secrets, one item per version: name, type and the repositories that get it. Never the values.
list_secretsList secretsGETLists the Cloud Agents secrets the API key can see (names, types, repositories and owner, never the values). A team admin's own user key also gets members' personal secrets, with their email.
list_team_secretsList team secretsGETLists the team's Cloud Agents secrets, one item per version: name, type (runtime secret, environment variable or build secret) and the repositories that get it. Never the values.

Active, 1 operation

get_active_environment_buildGet active environment buildGETTells what an agent started on this environment boots from: one of its builds (type build, with buildId) or Cursor's default image (type universal_image).

Archive, 1 operation

archive_agentArchive agentPOSTArchives an agent: it stays readable but accepts no new runs until unarchive_agent. Reversible, unlike delete_agent.

Artifact, 1 operation

list_artifactsList artifactsGETLists the files an agent saved under its workspace artifacts/ directory (at most 100), with path, size and last change. Pass a path to download_artifact for a download link.

Cancel, 1 operation

cancel_runCancel runPOSTCancels the active run of an agent; the run ends as CANCELLED. A run that already ended returns 409 run_not_cancellable. The agent stays usable: send a new prompt with create_run.

Download, 1 operation

download_artifactDownload artifactGETReturns a temporary download URL (valid 15 minutes) for one artifact of an agent, given its path from list_artifacts.

History, 1 operation

list_environment_historyList environment historyGETLists the changes made to an environment, newest first, as its dashboard History tab shows them: kind, title, details, source and the environment.json each change saved.

Me, 1 operation

get_api_key_infoGet api key infoGETReturns the name and creation date of the API key in use and, for a user key, its owner's Cursor user id, email and name (personal data). Useful to check the connection works.

Model, 1 operation

list_modelsList modelsGETLists the recommended models an agent can run with: id (pass it as model_id), name, aliases, the parameters each accepts and its valid variants.

Repository, 1 operation

list_repositoriesList repositoriesGETLists the GitHub repositories the user can reach through Cursor's GitHub App. Very strict rate limit (1 call per minute, 30 per hour) and it can take tens of seconds: call it rarely.

Unarchive, 1 operation

unarchive_agentUnarchive agentPOSTUnarchives an agent so it accepts new runs again.

Usage, 1 operation

get_agent_usageGet agent usageGETReturns the tokens an agent consumed (input, output, cache write, cache read, total), summed and per run, or for one run with runId. Early access: accounts without it get 403 feature_unavailable.

Use Cursor Cloud Agents from Claude Code, Cursor or Codex

Connect Cursor Cloud Agents in LiveContext, create an API key in Settings, then add the LiveContext MCP server to your client. Its Cursor Cloud Agents operations become tools the client can call, next to every other integration you connected.

Claude Code

claude mcp add --transport http livecontext https://livecontext.ai/mcp --header "X-API-Key: YOUR_API_KEY"

Codex (~/.codex/config.toml)

[mcp_servers.livecontext]
url = "https://livecontext.ai/mcp"

[mcp_servers.livecontext.http_headers]
"X-API-Key" = "YOUR_API_KEY"

Cursor, Claude Desktop and other JSON configs

{
  "mcpServers": {
    "livecontext": {
      "type": "http",
      "url": "https://livecontext.ai/mcp",
      "headers": {
        "X-API-Key": "YOUR_API_KEY"
      }
    }
  }
}

Set up the MCP server step by step

How to connect Cursor Cloud Agents

Create your workspace
Sign up free. Nothing to install.
Connect Cursor Cloud Agents once
Add your Cursor Cloud Agents connection in Settings. Every tool from this integration reuses it, and secrets are encrypted at rest and visible only to you.
Describe the job in chat
The workflow builds itself in front of you and picks the endpoints it needs. Drag the steps yourself on the canvas whenever you would rather be explicit.