Introduction

PromptJang Cloud provides reliable delivery and durable mailboxes for applications and agents. Producers use HTTP or the official MCP server. Every item goes to a registered webhook, mailbox, or A2A target and uses one acceptance, idempotency, retention, and billing contract.

พร้อมแจ้ง — Ready to Inform.

The Problem

Reliable message delivery is deceptively hard:

• Retries — endpoints go down, networks drop, servers return 500s. Events get lost silently.
• Trust — how does the receiver know the message really came from you? How do you prove delivery?
• Observability — did it arrive? What did the endpoint return? How long did it take?
• Recovery — your endpoint was down for 2 hours. Now what?
• Agents — an agent needs bounded tools, registered destinations, idempotency, and evidence when it triggers work outside the model.

Most teams build a half-baked retry loop and move on. The result: silent failures, angry customers, 3am pages.

The Solution

PromptJang is the delivery layer between application or agent producers and a registered target:

Three steps:

1. Create a target — start with an HTTPS webhook; add a mailbox or A2A agent when needed
2. Send JSON — one POST request with your API key, no signing required
3. Check status — query by event ID or your own correlation ID

Three operational stories

1. Webhook delivery — Application → PromptJang → service.
2. Agent mailbox — Application or a human-triggered system → PromptJang → agent through API or MCP.
3. A2A handoff — Agent → PromptJang → registered A2A agent.

The shared boundaries are:

• Registered targets only — MCP never accepts an arbitrary destination URL
• Signed payloads — the receiving service can verify the delivery came through PromptJang
• Idempotent ingestion — clients can reuse an Idempotency-Key safely when a request outcome is uncertain
• Target-specific handling — webhook and A2A push; mailbox consumers pull and claim
• Evidence — delivery attempts, mailbox lifecycle, or A2A acceptance are recorded

Every newly accepted webhook event, mailbox message, or A2A handoff creates one usage record. Retries, reads, claims, acknowledgements, replays, rejected requests, and idempotent duplicates create no additional usage. The published rate is $0.10 per 1,000 accepted items, but live billing is disabled during the founder-assisted beta.

After the founder provisions the account and the client creates an API key and target, runtime operations do not require a browser. The dashboard remains the control surface for API-key management, target registration, delivery evidence, local usage, and future billing access.

What PromptJang Does For You

• Retries with backoff — failed deliveries retry automatically after 60, 120, 240, 480, and 960 seconds
• Signs your webhooks — every outbound delivery uses Standard Webhooks v1, so your target can use established verification libraries
• Tracks every delivery — status code, latency, and error details for each attempt
• Event replay — re-deliver any past event with one API call
• Idempotency support — every event gets a unique ID your target can use to avoid processing duplicates
• SSRF protection — target URLs validated against private IPs and localhost
• Durable mailboxes — unread work survives consumer downtime and has an explicit claim owner
• A2A adaptation — standard outbound A2A v1 messages with acceptance evidence

PromptJang Relay is a separate Apache-2.0 product for reliable webhook delivery on your own PostgreSQL. It has its own API, documentation, releases, and container and does not include or require PromptJang Cloud mailboxes, MCP, A2A, organizations, or usage billing.

What You Don't Have To Build

Without PromptJang	With PromptJang
Retry queue with exponential backoff	One POST request
Signature generation and verification	We sign for you
Delivery attempt logging	Full audit trail in dashboard
Payload storage for replay	Stored automatically
Failure investigation UI	Attempt timeline in the dashboard
SSRF protection on target URLs	Built-in