Everything between a photo and a verified business action.
RunOCR combines visual AI, your business data, deterministic rules and human review on one canvas, so visual work runs end to end instead of stopping at extracted text.
See. Understand. Verify. Act.
Every workflow follows the same four stages. You choose the steps in each one.
Photos, scans, PDFs, video frames and web pages arrive from uploads, the API, webhooks, schedules or new records.
Vision steps return the fields you describe, typed and with a confidence score.
Lookups, tolerances, thresholds and duplicate checks compare the result with your data.
Records are created or updated, APIs called, people alerted, or a person asked to decide.
50+ configurable steps in seven categories.
No industry-specific black boxes. Combine general steps into exactly the process you run today.
Decide how a run starts.
- Image uploaded
- Video uploaded (frames sampled)
- Website captured
- Scheduled execution
- API request
- Webhook received
- New record
Turn pixels into typed fields.
- Extract text (OCR)
- Extract structured fields
- Detect objects
- Identify products
- Barcodes & QR codes
- Compare images
- Detect visual changes
Bring in your business context.
- Find existing record
- Search database
- Match customer or employee
- Query reference dataset
- Historical observations
- Retrieve API data
Deterministic, readable rules.
- If / Else and Switch
- Compare with tolerance
- Threshold validation
- Confidence validation
- Required fields
- Duplicate detection
Reasoning that produces data.
- Classify information
- Infer attributes
- Summarize findings
- Evaluate against instructions
- Detect anomalies
- Map to schema
Do something with the result.
- Create or update records
- Send email & notifications
- Call API
- Trigger webhook
- Generate report
- Create exception
People where judgement matters.
- Request approval
- Manual verification
- Accept or reject result
- Assign to a teammate
- Request more evidence
- Escalate exception
Save any group of steps, such as “Verify Employee Identity”, and drop it into other workflows.
A canvas built for real operations.
Drag steps onto the canvas, connect their outcomes and configure each one in the side panel. Branches, merges, error paths and reusable components are all first-class.
- Drag-and-drop canvas with branching
- Insert any earlier step's output
- Typed outputs you define per step
- Workflow-level input rules
- Debug mode on the canvas
- Versions with one-click rollback
- Retries, timeouts and error paths
- Component library
AI that proposes. Rules that decide.
AI is treated as a source of typed evidence. Business actions only happen through rules you can read, or through people you trust.
Every vision and AI step returns fields you define. Empty required fields fail the step, so nothing downstream acts on a guess.
Conditions are plain rules over typed values, so you can always see why a branch was taken.
Reviewers get a private link to approve, correct values or upload evidence. No account needed.
Publishing freezes a version. Live runs never change under your feet.
See every step's input, output, outcome and timing while a test run executes.
Each run keeps its inputs, decisions, reviewer actions and outputs.
Starts from anywhere. Reports back to anything.
Runs start from the app, your systems or a schedule, and results flow into your tools through APIs, webhooks, email and records.
Secret inbound URLs per workflow and HMAC-signed outbound calls.
Start runs with JSON or files and wait for the typed result.
Run every few minutes, hourly or daily in your time zone.
Full-page screenshots of any public page, on a schedule.
Keep reference data in RunOCR, or call your own API.
Email a person a link to decide, correct or add evidence.
Build your first visual workflow today.
Start from a template, upload a real photo and see each step's result in minutes.
No card needed · 3 workflows and 100 runs free · Cancel anytime