A2PCheck for AI agents & coding harnesses
A2PCheck for AI agents and coding harnesses. Validate an A2P 10DLC campaign against common Twilio / TCR rejection patterns before your agent submits it through a provider.
Guidance only — A2PCheck is independent and not affiliated with Twilio, The Campaign Registry, or any carrier, and we can't guarantee approval.
Agents should not submit A2P campaigns blindly
A rejected A2P 10DLC campaign costs days of carrier/TCR re-review and, with some providers, re-vetting fees. Before your agent calls a provider's registration API (Twilio, Telnyx, Vonage, Plivo, Bandwidth), have it validate the draft, read the per-field findings, remediate, and re-validate until the verdict is GREEN.
A2PCheck is independent — not affiliated with Twilio, The Campaign Registry, or any carrier. It does not submit campaigns and cannot guarantee approval; only the official review process approves a campaign. Send campaign metadata and public URLs only — never end-user phone numbers, customer records, API keys, or secrets.
Validate over HTTP
One JSON endpoint. tier defaults to basic; use full to crawl the website and policy URLs (slower). A bearer token is optional — without one you get the anonymous basic tier (rate-limited per IP).
curl -sS -X POST https://www.a2pcheck.com/api/agent/validate \
-H "Content-Type: application/json" \
-d '{
"tier": "full",
"campaign": {
"useCaseType": "MARKETING",
"campaignDescription": "Weekly promotional offers to opted-in customers...",
"sampleMessages": [
"Acme: 20% off this week. Shop acme.example.com/sale. Reply STOP to opt out.",
"Acme: New arrivals. Browse acme.example.com/new. Msg&data rates may apply."
],
"messageFlow": "Customers opt in via an unchecked SMS box at checkout."
}
}'The response carries overallTier (RED / YELLOW / GREEN), criticalIssues, warnings, and per-field fieldResults with fix suggestions. Full field reference is in /llms-full.txt.
Use the agent skill
A skill package ships in the source repo for harnesses that load skill folders (OpenClaw-style, Claude Code, Codex, Hermes). It encodes the validate → remediate → re-validate workflow and the campaign schema. Copy it into your skills directory:
# generic skill folder lives at skill/ in the repo cp -r skill ~/.agents/skills/a2pcheck
Harness-specific notes: OpenClaw and Claude Code. The skill is not yet published to a skills registry — use the in-repo copy for now.
Machine-readable surfaces
- /llms.txt — index for AI assistants.
- /llms-full.txt — full agent reference: schema, routes, response fields, caveats.
- /openapi.json — OpenAPI 3.1 contract for the validation endpoint.
- /api-docs — human-readable API reference.
On the roadmap
A hosted MCP server and a CLI are planned so agents can discover and call A2PCheck without hand-wiring HTTP. Neither is published yet — for now use the HTTP endpoint and the skill above.