Headless forms
One endpoint. Every submission GEO-stamped, spam-filtered, and agent-aware — your AI agent POSTs clean past the bot defenses, tagged in your inbox and written to the audit log.
→ call it liveThree things your agents need, one backend: form endpoints they POST to and sail past the bot defenses, real mailboxes they send, receive, reply, and read OTP codes from, and disposable inboxes for testing email in CI. One token system, one OpenAPI spec they read themselves.
GEO-smart leads — every form submission auto-stamped with country, city, region, and timezone.
Try the live demo on the right → submit, watch the geo light up.
Or call the API live ↓ — all three capabilities, real endpoints.
Forms in, mail out — one backend, one token system, one console. Each surface is in the OpenAPI spec your agent reads itself.
One endpoint. Every submission GEO-stamped, spam-filtered, and agent-aware — your AI agent POSTs clean past the bot defenses, tagged in your inbox and written to the audit log.
→ call it live
Real email inboxes your agents own: pick an identity
(support@agent.ollastack.com), then send, receive,
and reply over the API — inbound is spam-filtered, so the inbox
stays clean.
Disposable inboxes on @test.ollastack.com:
/wait for the email, read codes[0] (the
OTP), assert. Unfiltered — a test sees every message.
React to what comes in, triage it with AI, and run it as a team — without standing up a single consumer service.
Auto-tag a lead, route an urgent enquiry to Slack or email, auto-reply to an invoice request — rules run server-side the moment a submission or email lands.
→ /docs/workflow-rulesCategory, sentiment, urgency, a one-line summary, and a suggested reply on every lead and email — on demand, as a rule action, or auto-run on each inbound message.
→ /docs/workflow-rulesInvite teammates into a shared workspace with the right access — owner, admin, member, or a read-only viewer for clients and stakeholders.
→ /docs/teamsGroup forms and agent mailboxes by client, product, or environment — one tidy workspace per project, with file uploads handled by presigned storage.
→ /docs/projectsForms in, mail out, inboxes your agents own. The Forms and Test-inbox demos run for real, right here — no key, no signup. The Agent Mail send is the exact endpoint; copy it into your terminal with a token.
$ curl -X POST https://login.ollastack.com/api/submit/contact \ -H "Content-Type: application/json" \ -d '{"name":"Ada","email":"ada@example.com","message":"Tell me more"}'
# the response shows here — geo (country / city / lat-long) is
# auto-stamped server-side from Cloudflare visitor headers Full reference + 3 SDKs in the docs — the whole surface is in /api/openapi.json, CORS-open so your agent reads it itself.
Same URL serves your AI agent, your static HTML form, your React app, your mobile client. Pick a tab.
$ curl -X POST https://login.ollastack.com/api/submit/contact \ -H "Authorization: Bearer fmd_<your-agent-token>" \ -H "Content-Type: application/json" \ -d '{"name":"Ada","message":"hi"}' {"success": true, "id": "sub_8f2…", "country": "IN", "city": "Mumbai"} # tagged in the inbox with the agent's label # every call recorded in the audit log # geo auto-attached — country / city / region / lat-long / timezone
Connects to Gmail, Sheets, Notion, Slack, HubSpot — and 1,000+ more via Zapier, Make, n8n, Activepieces — /docs/integrations.
Three small steps. One curl block. Your agent is sending real submissions before your coffee cools.
Sign up, click Create form, name it. Your endpoint URL is yours immediately — no review, no approval queue.
→ create your first formOne click on Connect your agent mints a scoped, audited token and hands you a paste-ready agent prompt plus a curl/SDK quickstart. No token plumbing to write yourself.
→ connect your agentOne Bearer-authed POST and the submission lands in your inbox, tagged with the agent's label, with every call written to the audit log.
→ /docs#fetchDesigned for AI-agent traffic from the ground up. Almost the whole surface is yours from your first token — sending from your own domain is the one paid-plan extra.
ollastack is a developer surface, not a SaaS dashboard your customers log into. You hand your agent the API key and the OpenAPI spec, and it handles every form action — submit, read, tag, export, configure — through one endpoint. The dashboard is a thin console for token management and inbox triage. That's the entire product, on purpose.
Every tier includes the same core features — the monthly submission cap is the main difference (sending from your own domain is a paid-plan extra). No overage charges — if you hit the cap, submissions stop until next month or you upgrade.
$0 / month
For trying it out and side projects.
$9 / month
For developers shipping their first agent-backed form.
$29 / month
For teams running multiple agent integrations at volume.
The short version of what ollastack is and how agents use it.
An agent email API lets an AI agent send, receive, and read email over HTTP instead of SMTP. Ollastack gives each agent a real mailbox (handle@agent.ollastack.com) it controls with a scoped Bearer token — sending, threaded replies, inbound polling, and OTP and link extraction, all as REST calls documented in one OpenAPI spec.
Create a disposable inbox, trigger the email, then poll the mailbox's /wait endpoint until it arrives — codes[0] is the extracted OTP. Test inboxes on test.ollastack.com are never spam-filtered, so a CI job sees every message it triggers. It's a drop-in replacement for Mailosaur or Mailinator.
Yes — the Free plan includes 50 submissions per month and every core feature, with no credit card. Paid plans are $9/month (Solo, 500 submissions) and $29/month (Team, 5,000 submissions); the only paid-only extra is sending from your own verified domain.
Ollastack is a form backend and an agent-email platform in one. Beyond Formspree-style form endpoints, it gives your AI agents real mailboxes and disposable test inboxes, Bearer tokens that skip the bot defenses, and a spam pipeline that quarantines a suspicious lead rather than silently dropping it.
Yes. A Bearer token owned by the form owner submits as a trusted agent — skipping honeypot, CAPTCHA, and per-IP rate limits — and is stamped in the inbox with the agent's label. The same token system sends and reads agent mail, and every call is written to an audit log.
Yes — persistent agent mailboxes run inbound mail through the same layered spam pipeline as forms (honeypot, blocklist, keyword and regex rules, Akismet, and an ML classifier), and it fails open so a classifier error never drops real mail. That spam-filtering is the main differentiator from AgentMail; disposable test inboxes are deliberately left unfiltered.
Mint a token, drop in a fetch call, ship. High-volume or custom needs? Talk to us.