Validate manifests, catch action drift, and automate readiness checks.

ACP Watchtower is the fastest way to turn an ACP manifest into a clear readiness score, a release-risk diff, and a shareable report your team can actually act on.

Readiness scoringAction lintingManifest diff modeWebhook-ready reportsScheduled re-checks

Build status

Tonight’s MVP stack is live, demoable, and automation-ready.

MVP LIVE

✓Validator workbench with manifest paste and demo samples

✓Readiness scoring with severity cards, category subscores, and fix guidance

✓Diff mode for action additions, removals, parameter changes, and description drift

✓Hosted report generation with file-based persistence and shareable URLs

✓Webhook stub + Vercel Cron-ready scheduled re-check endpoint

Manifest validation

Run deterministic structural checks, metadata linting, and readiness scoring before ACP apps hit production.

Release diffing

Compare old vs new manifests, flag risky action drift, and generate a short release-risk summary.

Automation hooks

Persist hosted reports, emit webhook-ready payloads, and expose scheduled re-check endpoints for recurring monitoring.

Automation-first revenue path

Free validator now, monitoring and alerts next.

• Free: manual validation, readiness scoring, and basic release diffing
• Pro: saved history, nightly re-checks, webhook alerts, and score trends
• Design partner: rollout reviews, report customization, and reliability audits

What ships tonight

1. Deterministic validator engine
2. Score + issues + recommendations UI
3. Diff mode for release-risk reviews
4. Hosted report persistence and public report pages
5. Webhook schema + Vercel Cron monitoring hooks

Get early access to Pro

Scheduled re-checks, webhook alerts, team dashboards, and CI/CD integration. Join the waitlist — we launch soon.

Validator workbench

Analyze a manifest in one pass

{
  "name": "Acme Support Agent",
  "version": "0.9.0",
  "description": "Agent-facing support toolkit for triage and escalation.",
  "metadata": {
    "owner": "support-platform",
    "contact": "agents@acme.dev"
  },
  "actions": [
    {
      "id": "search-kb",
      "title": "Search knowledge base",
      "description": "Search the support knowledge base using a customer issue query.",
      "input": {
        "properties": {
          "query": { "type": "string", "description": "Customer problem or keywords" },
          "limit": { "type": "number", "description": "Number of results to return" }
        },
        "required": ["query"]
      },
      "examples": [{ "query": "refund policy for annual plan" }]
    },
    {
      "id": "create-escalation",
      "title": "Create escalation ticket",
      "description": "Create an escalation ticket for incidents that need a human support lead.",
      "input": {
        "properties": {
          "accountId": { "type": "string", "description": "Customer account id" },
          "reason": { "type": "string", "description": "Why this escalation is needed" }
        },
        "required": ["accountId", "reason"]
      },
      "examples": [{ "accountId": "acct_123", "reason": "billing lockout" }],
      "safetyNotes": ["Escalations are logged and visible to the support lead queue."]
    }
  ]
}

Automation layer

Nightly monitoring-ready

• POST `/api/analyze` for deterministic manifest scoring
• POST `/api/diff` for release-risk checks before deploys
• POST `/api/reports` to persist a hosted report snapshot
• POST `/api/webhooks/ingest` as the results delivery stub
• GET `/api/cron/recheck` for Vercel Cron-based scheduled monitoring

Good default loop: re-check tracked manifests nightly, diff against previous version, then ship the webhook payload to Slack/Discord/internal queues.

Readiness output

Score, severity, and fixes

Paste an ACP manifest above and hit “Run readiness analysis” to see severity cards, category scores, and issue-level fix guidance.

Diff mode

Release risk before shipping

{
  "name": "Acme Support Agent",
  "version": "1.0.0",
  "description": "Agent-facing support toolkit for triage, escalation, and renewal interventions.",
  "metadata": {
    "owner": "support-platform",
    "contact": "agents@acme.dev",
    "docsUrl": "https://example.com/docs/support-agent"
  },
  "actions": [
    {
      "id": "search-kb",
      "title": "Search knowledge base",
      "description": "Search the support knowledge base with semantic matching and return concise snippets.",
      "input": {
        "properties": {
          "query": { "type": "string", "description": "Customer problem or keywords" },
          "limit": { "type": "number", "description": "Number of results to return" },
          "locale": { "type": "string", "description": "Optional customer locale" }
        },
        "required": ["query"]
      },
      "examples": [{ "query": "refund policy for annual plan", "locale": "en-US" }]
    },
    {
      "id": "create-escalation",
      "title": "Create escalation ticket",
      "description": "Create an escalation ticket for incidents that need a human support lead, including severity and customer impact.",
      "input": {
        "properties": {
          "accountId": { "type": "string", "description": "Customer account id" },
          "reason": { "type": "string", "description": "Why this escalation is needed" },
          "severity": { "type": "string", "description": "incident severity level" }
        },
        "required": ["accountId", "reason", "severity"]
      },
      "examples": [{ "accountId": "acct_123", "reason": "billing lockout", "severity": "high" }],
      "safetyNotes": ["Escalations are logged and visible to the support lead queue."]
    },
    {
      "id": "draft-renewal-offer",
      "title": "Draft renewal offer",
      "description": "Draft a renewal offer suggestion for at-risk accounts before routing to a human approver.",
      "input": {
        "properties": {
          "accountId": { "type": "string", "description": "Customer account id" },
          "riskReason": { "type": "string", "description": "Why the renewal looks at risk" }
        },
        "required": ["accountId", "riskReason"]
      },
      "examples": [{ "accountId": "acct_991", "riskReason": "downgrade signals" }],
      "safetyNotes": ["Draft only. Human approval required before customer send."]
    }
  ]
}

Webhook preview

Event payload

{
  "message": "Create a report to preview the outbound event schema."
}

FAQ

Common questions

What is ACP?

The Agent Communication Protocol defines how AI agents discover and invoke actions exposed by applications. Think of it as an API contract designed specifically for agent-to-app communication.

Do you store my manifest?

Only if you explicitly create a hosted report. Ad-hoc analyses are stateless — nothing is saved unless you click 'Create hosted report'.

How is the readiness score calculated?

We run deterministic checks across five categories: schema completeness, action clarity, discoverability, safety guardrails, and maintainability. Each issue reduces the category score based on severity.

Can I automate this?

Yes. POST to /api/analyze with your manifest JSON, or set up /api/cron/recheck for scheduled monitoring. GitHub webhook support is also available for CI/CD integration.

Is there a CLI?

Not yet — but the API is fully usable from curl or any HTTP client. A CLI is on the roadmap for Pro users.

What does the badge show?

After creating a hosted report, you get an SVG badge URL showing your readiness score and verdict. Embed it in your README to signal agent-readiness to your users.