Support
Tell us what you need
We will route it to the right place. Most questions are answered in the docs before a person is needed, so that door is first.
Self help
The answers are already public
The quickstart, every MCP tool, every REST route, every error code and every rate limit. If your question starts with “how do I”, it is probably answered here.
Three steps and a restart.
The whole reference set.
Transports, and every tool.
Send, lease, ack, redeliver.
Scopes, rotation, revocation.
Every POSTMQ_* code.
Limits, headers, backoff.
Posture, and what is not built.
Or tell a person
Pick the closest match — the category decides which mailbox reads it, and you can change it on the form.
Something is broken, or an idea
A call that errors, a message that never arrives, a lease behaving strangely — or something you wish an agent could do. An error code or a message id shortens the round trip.
Account, billing & usage
Sign-in trouble, workspace questions, and anything about your plan or allowances. The pricing page carries the current tiers; this is for everything it does not answer.
Sales, partnerships & press
Enterprise evaluation, a DPA request, a partnership, or a press question. The security questionnaire is already public if that is what you came for.
Privacy & your data
Ask about access, correction, erasure, export, or workspace deletion — for your own account, or for personal data of yours that reached us some other way. You do not need an account to ask.
Report content
Content carried by PostMQ that you believe is illegal, breaks the acceptable-use policy, or infringes
copyright. Anyone can report — you do not need an account. If you would rather write to us, email
[email protected], or [email protected] for a copyright complaint. Every
report that meets the required shape is reviewed by a person.
Security disclosure
Found a vulnerability, or suspect a credential is compromised? Email
[email protected] — the address is public on purpose, so nothing sits between you and us.
We acknowledge within three business days. There is no bug bounty, and we would rather say so than leave it
ambiguous.
Frequently asked
Durable messaging and session state for AI coding agents. Agents hand work to each other as typed, durable messages, and keep a session record people can read. The home page has the short version; the docs have the whole surface.
A file tells you nothing about whether anyone picked it up. A handoff is delivered under a lease and either acknowledged, declined, redelivered or dead-lettered, and neither agent has to be running when the other is. The handoffs page walks the round trip.
It waits in that agent’s inbox until the message’s delivery time-to-live lapses. Sending commits atomically: the envelope, its idempotency record and its audit row go in one transaction, so a message either fully exists or does not exist at all.
Messages are claimed under a lease. If the holder neither acks nor nacks before it expires the message returns to the queue and is redelivered; past the per-priority attempt ceiling it goes to dead-letter instead of retrying forever. Delivery is at-least-once, so make handlers idempotent on the message id.
Create an AI account, issue it a credential, and configure your MCP client. The quickstart walks it end to end, and the integrations pages carry the per-client configuration.
PostMQ speaks standard MCP, so any MCP-capable client can connect. The integrations pages list the clients we document setup for.
If you lost it, rotate: credentials are once-disclosed — the plaintext is shown at issue and never again — so rotation is the recovery path. If you suspect compromise, revoke or rotate immediately and then tell us at [email protected] so we can check for misuse.
There is a free tier. The pricing page carries the current tiers, what each includes, and what is and is not open yet — it is the one place those numbers live, so we do not restate them here and let the two drift.
Yes, today, over the versioned REST API: every ledger — sessions, decisions, backlog, lessons, rules, usage — is readable on list and query routes, so a script can pull it. There is no one-click export yet.
Use the “Something is broken” card above. Include what you called, what you expected, and what happened — an error code, a message id or a timestamp turns a guess into a lookup. Please do not paste credentials or message payloads; describe them instead.
Email [email protected] with enough detail to reproduce. We acknowledge within three business days, and there is no bug bounty — we would rather say so than leave it ambiguous. The security page carries the scope and the safe-harbour terms. Please do not open a public issue for a security report.
The pricing page is where response commitments live, by tier. Two exist today and are kept deliberately narrow: security reports are acknowledged within three business days, and every content report that meets the required shape is reviewed by a person.