metaautomationscursorcloud-agentsUpdated 2026-07-14

Cursor Automations Setup

Cursor Automations are the cloud runner for wiki jobs that can operate from a fresh repo clone plus MCP tools.

Routing

Open this page when creating, auditing, repairing, or explaining Cursor cloud automations. Use Operational Heartbeat and npx tsx scripts/check-freshness.ts for run freshness; use this page for setup shape, cloud constraints, and which jobs should stay local.

As of 2026-07-14, the repo has 22 canonical automation definitions under automations/; 14 have cloud-adapted .cursor-automation.md prompts. Local-only jobs need filesystem credentials, browser sessions, local transcripts, or cost-gated manual review. Source: find automations, 2026-07-14

Cloud-Ready Automations

These have <slug>.cursor-automation.md files and can be configured at cursor.com/automations:

Slug Schedule Primary MCP / source
career-pipeline daily Firecrawl / job boards
code-bugfix daily GitHub
content-pipeline daily Firecrawl / social
daily-work-summary daily GitHub / transcripts when available
frontend-frontier-radar daily Firecrawl
language-orchestration-radar weekly Firecrawl / papers
oss-license-audit weekly GitHub
project-briefing daily GitHub
recruiter-discovery weekly Firecrawl
self-heal on-demand GitHub
seo-geo-radar weekly Firecrawl
signal-radar daily Firecrawl
source-compile every 4h locally; daily-compatible in cloud repo-committed raw sources

Local or Manual Automations

Slug Why not normal cloud
calendar-sync Google credentials / local integration setup
email-sync Gmail credentials / local integration setup
slack-sync Slack browser session / local agent-browser setup
meeting-sync Circleback token / local integration setup
cross-project-learning local transcripts across projects
skill-maintenance local skill directories and recent sessions
security-scan cost-gated, target-repo-specific, manual review

Setup Flow

For each cloud-ready automation:

  1. Open cursor.com/automations.
  2. Create a new automation.
  3. Set repository to kevin-wiki.
  4. Set the schedule from the automation frontmatter.
  5. Enable required MCPs: Firecrawl for web scans, GitHub for repo data.
  6. Paste the instructions from automations/<slug>.cursor-automation.md.
  7. Run once manually if the UI supports it.
  8. Confirm a PR or output file appears.
  9. Merge the PR so state.json updates freshness.

Cloud Runner Constraints

Cursor cloud agents run from a fresh sandbox. They cannot read:

  • ~/.cursor/projects/ transcripts
  • ~/.codex/sessions/
  • local browser profiles
  • local OAuth tokens
  • uncommitted files on Kevin's machine

If an automation needs those, run it locally or commit the raw source first.

Output Contract

Cloud automations should:

  1. read automations/_cursor-automation-preamble.md
  2. write outputs/<YYYY-MM-DD>/<slug>/cursor-automation.md
  3. update wiki targets only when evidence supports it
  4. update state.json
  5. append wiki/log.md
  6. run generated-surface closeout commands from Generated Surface Contract when generated packs, registries, redirects, qmd, or stateful ledgers changed
  7. open a PR

Monitoring

Use:

npm run audit:automations
npx tsx scripts/check-freshness.ts

The readiness audit is authoritative for installed/configured prerequisites and live Codex schedule drift. The freshness script is authoritative for overdue and never-run status. Cursor's web UI is the source for cloud run history and cloud MCP enablement.

Timeline

Related Pages

Automation SchemaScheduling Operating ModelAutomation ReadinessAutomation Proof ContractGenerated Surface ContractOperational HeartbeatSource Compile Automation SetupSlack Sync Workflow

4 pages link here

Agent Config SyncMetaAutomation ReadinessMetaAutomation SchemaMetaNo One-Off WorkPhilosophies