Skip to content
OutcomeOS
Get Started

Open source · runs on Cloudflare's edge

Automate outcomes. Not just workflows.

Your business doesn't need more automations. It needs results. Connect your applications, orchestrate business processes and AI agents, and track every outcome from execution to completion.

Start Building Free Explore the Platform

Free Starter plan · no card · or self-host it

app.outcomeos · Outcomes / Lead responsePreview · sample data

Respond to every qualified lead in 5 min

Active

Achieved when the customer replies — not when we send. Verified by inbound reply events.

42%

outcome achieved

Leads received

100

this week

Contacted < 5 min

96

96% of leads

Replies verified

42

outcome achieved

Need attention

4

4 failed executions

Workflow · v3 published

Webhook

New lead

WhatsApp

Send first

Wait

Reply, 24h

Replied

Update CRM

Notify sales

No reply

Send SMS

Fallback

WebhooksCRM MessagingTwilioHubSpot

Recent executions

  • Priya S. ex_8f2a

    Reply in 3m 40s

  • Marco D. ex_8f29

    Delivered, no reply yet

  • Aisha K. ex_8f27

    WhatsApp failed: token expired

  • Tom R. ex_8f24

    SMS rejected: invalid number

Median first response 2m 14s · 54 follow-ups pending

OutcomeOS is an outcome-based automation platform for growing businesses, agencies and SaaS teams. You define a business result, connect your apps, and OutcomeOS runs the workflow and verifies the result with real events. The Starter plan is free, and the code is open source so you can run it yourself.

The problem

Your workflows run. But are your business goals being achieved?

01

Execution doesn't guarantee results.

“Workflow succeeded” means the steps ran. It doesn't mean the lead answered or the invoice got paid.

02

Failed integrations fail quietly.

A token expires, an API returns 500, and the process stops. You find out from a customer.

03

Nobody can see the impact.

Run counts and task usage don't tell you how many customers were contacted, onboarded or recovered.

How it works

One platform. From intent to outcome.

A workflow is an implementation detail. The business result is the product.

  1. 1

    Define the outcome

    Say what success looks like: “every qualified lead replies within 24 hours”. Set the target and the deadline.

  2. 2

    Connect your apps

    Start with a signed webhook and HTTP. Add messaging, CRM and payments as connectors. Connections are tested before they count.

  3. 3

    Orchestrate execution

    Build the workflow visually. Steps run from a queue with retries, waits and idempotency keys.

  4. 4

    Measure the result

    Events from the real systems mark each outcome achieved, at risk or failed. Every number links to its source.

Start Building Free

The platform

Six parts. One job: get the work done.

Outcome Management

Each objective has success criteria, a target and a deadline. Progress counts verified results only.

OutcomesPreview · sample data

  • Lead response < 5 min

    Sales · 87 / 100 leads

    On track

  • Recover failed payments

    Revenue · $3,240 verified

    At risk

  • Onboard in 48h

    Success · 16 / 25 customers

    On track

  • Resolve tickets in 3 days

    Support · 11 / 50 issues

    Failed: 2

Visual Workflow Builder

Triggers, actions, waits and branches on a canvas. Every published version is immutable.

Lead response · Workflow editorPreview · sample data

Webhook received

Trigger

Send WhatsApp

Selected

If replied within 24h

Condition

Update CRM

Send SMS

Send WhatsApp

Connection
CRM Messaging
To
{{trigger.phone}}
Retry policy
No blind retry

Valid

Integration Marketplace

Connectors share one SDK. A connection shows “connected” only after a live test passes.

IntegrationsPreview · sample data

  • Webhooks

    Connected · tested

  • HTTP Request

    Connected · tested

  • CRM Messaging

    Not connected

  • Twilio

    Test failed: 401

  • WhatsApp Business Platform

    Not connected

  • Gmail

    Not connected

Intelligent Recovery

Safe steps retry with backoff. Unsafe ones stop and tell you why. Replay when the cause is fixed.

Execution ex_8f27 · step timelinePreview · sample data

  1. webhook.received

    Completed

    12ms

  2. http.request · enrich lead

    Completed after 2 retries

    503 → 503 → 200 · backoff 2s, 4s

  3. whatsapp.send

    Failed · not retried

    401 token expired · non-idempotent

  4. sms.send (fallback)

    Completed

    idem_key lead_381:sms

Secrets masked in logs: Bearer ••••••3f9a Replay

Communications Automation

WhatsApp, SMS and email as first-class steps, with delivery status, replies and channel fallback.

Customer timeline · Aisha K.Preview · sample data

  • WhatsApp

    Hi Aisha, thanks for asking about a demo. Want to pick a time?

    Failed · connection expired

  • SMS

    Hi Aisha, it's Sam from Acme. Want to pick a demo time? Reply here.

    Delivered

  • SMS

    Yes — tomorrow at 10 works.

    Reply received · outcome verified

CRM updated · Sales notified

AI Workflow Assistant

Describe the goal in plain words and get a draft workflow to review. Early preview: drafts only, you approve every publish.

AI workflow assistant Early previewPreview · sample data
Every new website lead gets a WhatsApp. No reply in 24h? Send an SMS. If they reply, update Zoho and tell sales.

Proposal: Lead response

  1. Trigger: webhook from your site
  2. Send WhatsApp via CRM Messaging
  3. Wait 24h for a reply → else SMS
  4. On reply: update Zoho CRM, notify sales

Zoho CRM isn't connected yet.

Review draftEdit

Nothing publishes or runs until you approve it.

See the whole platform

Use cases

Start with the outcome that costs you most.

Integrations

Honest about what's connected.

2 integrations are available today: signed webhooks and HTTP requests, which reach any REST API. The rest are labelled beta, needs setup or coming soon — and we never show a connection as working until a live test passes.

Browse integrations

Reliability

When automations fail, your business shouldn't stop.

Runs are queue-driven with durable state, so a long wait or a flaky API doesn't lose work. Every failure has a reason you can read.

  • Retries with exponential backoff

    Configurable per step. Timeouts and server errors retry; errors that won't fix themselves don't.

  • Duplicate prevention

    Every run and every external call carries an idempotency key. The same trigger twice runs once.

  • No blind retries

    Payments and outbound messages are never retried when the result is unknown. They go to a human.

  • Execution monitoring

    Every run has an ID. Every step records input, output, timing, retry count and error.

  • Error recovery and replay

    Fix the connection, then replay the failed run. The original stays in history.

  • Detailed, masked logs

    Full step logs with tokens and secrets masked before they're stored.

  • Dead-letter queue

    Runs that exhaust their retries land in one queue with the reason attached.

  • Read how it works in the docs.

    Reliability docs

For AI agents

Your agent can run it too.

OutcomeOS ships an MCP server at /mcp and a REST API. Give your assistant an API key and it can create outcomes, publish workflows, run them and replay failures — inside your permissions.

MCP docsDeveloper overview
Assistant · connected to OutcomeOS MCPExample conversation
  • What outcomes are behind this week?
  • list_outcomes“Recover failed payments” is at risk: 41% of target with 3 days left.
  • Set up lead response for the new landing page. WhatsApp first, SMS after 24 hours.
  • create_outcome · create_workflowCreated the outcome and a draft workflow. Want me to publish it?
  • Yes, publish it and send a test lead.
  • publish_workflow · run_workflowPublished v1. Test run ex_91c0 is waiting for a reply.
  • Any failures overnight?
  • list_executions · replay_execution2 failed on an expired token. You reconnected it, so I replayed both. Both completed.
  • ingest_outcome_eventYour CRM reported a booked call for lead_381. I recorded it against the outcome.

Why I'm building this

Hi, I'm [Founder name].

I've built a lot of automations that said “success” while the lead never heard from us. The workflow ran. The business result didn't happen. Nobody noticed for days.

So I'm building the tool I wanted:

  1. 1

    Outcomes come first. You say what result you want. Workflows are how you get there, and they're judged by that result.

  2. 2

    Failures get a reason and a fix. Retries where they're safe, a human where they're not, and a replay button for the rest.

  3. 3

    It's open source. Run it on your own server or your own Cloudflare account. The hosted plans are there if you'd rather we run it.

Read the code · About OutcomeOS

Pricing

Free to start. Pay when it's doing real work.

Plans are priced on active workflows and executions. Every plan gets outcomes, the workflow builder and execution logs.

  • Starter

    For trying OutcomeOS on one real process.

    3 active workflows · 1,000 executions/mo

  • Growth

    For teams running revenue-critical automations.

    25 active workflows · 25,000 executions/mo

  • Business

    For agencies and companies with many processes.

    250 active workflows · 250,000 executions/mo

See plans and what's included ·Compare alternatives

FAQ

Questions people ask before switching.

How is this different from Zapier, Make or n8n?
Those tools are great at running steps; OutcomeOS also tracks whether the business result happened. You define an outcome with success criteria, like “the lead replied”. Workflows send events toward it, and the outcome is only marked achieved when a verifying event arrives. Run status and outcome status are shown separately.See the comparison →
How do you verify an outcome was actually achieved?
With events from the system where the result lives, not from the workflow itself. A lead outcome needs an inbound reply event. A payment outcome needs a signed payment-succeeded event for the same customer and invoice. Each metric links back to the event and source that verified it.
What happens when an integration fails?
Safe steps retry with exponential backoff; unsafe ones stop and ask. Steps marked idempotent retry automatically. Sending a message or taking a payment is never blindly retried — it uses an idempotency key, and if the result is unknown the run goes to the attention queue with the reason. You can replay failed runs once the cause is fixed.
What happens when I hit my plan limit?
Your data stays readable, but new executions stop until the next cycle or an upgrade. Dashboards, logs and outcomes keep working. Triggers that arrive over the limit are rejected with a clear error rather than silently dropped, so the sender can retry later.Pricing →
How do AI agents connect?
With an API key, over the REST API or the MCP server at /mcp. Create an API key in Settings and pass it as a Bearer token. The MCP server exposes tools like create_outcome, create_workflow, run_workflow and replay_execution, scoped to your organization. Agents can't bypass your permissions.MCP docs →
Can I host it myself?
Yes. The code is open source and you can run it on your own server or your own Cloudflare account. The hosted plans are the managed service. Self-hosting is free; you bring your own provider credentials.Source code →Self-host guide →
Which integrations work today?
Webhooks and HTTP requests are available now; others are marked beta, needs setup or coming soon. We never show an integration as connected until a live test passes. With HTTP you can reach any REST API, including tools we don't have a connector for yet.Integration directory →
Where is my data stored, and are credentials safe?
Data runs on Cloudflare's network, and provider credentials are encrypted with AES-GCM before they're stored. Secrets are never sent back to the browser, and values that look like secrets are masked in execution logs. Every record is scoped to your organization.Security →
Do you offer refunds?
[Refund policy set by owner before launch.] The Starter plan is free, so you can test a real process before you pay.

One short letter when we ship something worth switching for.

Product updates, reliability notes and real automation patterns. No spam, no sharing your address.

Double opt-in. Unsubscribe from any issue in one click.

Latest: Issue 1 — Why we launched OutcomeOS

Start automating what actually matters.

Define one outcome, connect a webhook, and see what got done. Free to start, no card needed.

Create Your Free Account See pricing