End-to-end automation for your hardest document workflows.

Process documents at scale — with configurable business rules, cross-document checks, and human review for exceptions.

Start free Get a demo

For document workflows

To turn files into data

Automated by agents

Workflows

Build durable document pipelines that keep every parse, split, extraction, validation, and review step in one place.

Retab orchestrates the full document lifecycle without stitching brittle tools together.

Features
Workflow Builder
Human Review
Schema Versioning
Run Inspection

Features

Turn PDFs, scans, spreadsheets, and emails into typed data with APIs that understand layout, tables, sources, and schemas.

/parse

Acme Corp12 Market St, Boston MA

INVOICE

#2042

BILL TONorthwind Trading Co.88 Harbor Rd, Seattle WA

IssuedMar 4, 2024

DueApr 3, 2024

TermsNet 30

ITEMDESCRIPTIONAMT

A-1Widget assembly$480

A-2USB-C cable$96

A-3Mount bracket$58

Subtotal$634

Tax$57

Total$691

Authorized signature

Date

| Item | Description | Amt |

|---|---|---:|

| A-1 | Widget assembly | $480 |

| A-2 | USB-C cable | $96 |

| A-3 | Mount bracket | $58 |

| Subtotal | $634 |

| Tax | $57 |

| Total | $691 |

| Authorized signature | Date |

/extract

/edit

/split

/partition

/classify

Developers unlock mission-critical data from documents that were previously too complex to automate.

Features
Extract
Split
Edit
Parse
Classify

Reliability

Know when automation can run straight through, when it needs review, and exactly where every answer came from.

k-LLMs Consensus

Extraction Avalue 1

Extraction Bvalue 1

Extraction Cvalue 2

Likelihoods0.67%

Extraction A

{
  "loan_amount": 250000,
  "applicant_name": "Jane M. Doe",
  "debt_to_income": 0.34,
  "term_months": 360,
  "annual_income": 184000,
  "credit_score": 742,
  "property_value": 325000,
  "down_payment": 75000
}

Extraction B

{
  "loan_amount": 249500,
  "applicant_name": "Jane Marie Doe",
  "debt_to_income": 0.35,
  "term_months": 360,
  "annual_income": 181500,
  "credit_score": 739,
  "property_value": 324500,
  "down_payment": 75000
}

Extraction C

{
  "loan_amount": 250000,
  "applicant_name": "J. M. Doe",
  "debt_to_income": 0.33,
  "term_months": 359,
  "annual_income": 184000,
  "credit_score": 741,
  "property_value": 325000,
  "down_payment": 76000
}

Likelihoods

loan_amount96%

applicant_name99%

debt_to_income91%

term_months94%

annual_income93%

credit_score98%

property_value95%

down_payment97%

Evals & monitoring

0.0%

(avg)

0.00%

Accuracy

100%91%85%79%

Reviews

Review-based validation

invoice_total

0.41

Extraction result

invoice_total

$12,089.000.41

vendor_name

Acme Industrial0.97

due_date

2025-03-150.94

Operations teams approve exceptions with confidence scores, source grounding, evals, and audit-ready traces.

Features
Confidence Scoring
Source Grounding
Evals
Reviews

Agents

Retab exposes the full document workflow context, letting agents inspect runs, trace failures, and fix automations without leaving their loop.

Terminal

Claude Codev2.1.3Opus 4.7

> Publish the claims packet workflow after a final run check

●Let's inspect the claims packet workflow before publishing this draft.

↳workflows_get{ "workflow_id": "wf_claims_packet", "version": "draft" }

⎿5 blocks · split packet · extract ACORD · extract reports · review exceptions

●The graph shape is right. I want a cross-document check before clean claims go straight through.

↳workflows_blocks_create{ "workflow_id": "wf_claims_packet", "type": "conditional", "label": "policy_match" }

⎿✓policy_match · ACORD policy number must match police report

●Wiring the guard after the ACORD and police report extractions.

↳workflows_edges_create{ "source": "extract_acord", "target": "policy_match" }

⎿✓extract_acord -> policy_match

↳workflows_edges_create{ "source": "extract_police_report", "target": "policy_match" }

⎿✓extract_police_report -> policy_match

●Running the recent packet through the draft to see whether the guard catches the known mismatch.

↳workflows_runs_create{ "workflow_id": "wf_claims_packet", "version": "draft", "inputs": { "start": "claim_203.pdf" } }

⎿run_42c9 · split complete · 3 extraction branches running

↳workflows_runs_get{ "run_id": "run_42c9" }

⎿✗awaiting_review · policy mismatch · ACORD P-8947 · police P-8941

●Good catch. Source boxes show the police report value is handwritten, so this should route to review.

↳files_get_sources{ "extraction_id": "ext_police_203", "path": "policy_number" }

⎿✓pdf_bbox { page: 2, left: 0.58, top: 0.21, width: 0.12 }

↳workflows_blocks_update{ "block_id": "policy_match", "on_fail": "claims_review" }

⎿✓policy_match failures route to claims_review · clean packets continue

●One clean dry run before publish.

↳workflows_runs_create{ "workflow_id": "wf_claims_packet", "version": "draft", "inputs": { "start": "claim_clean.pdf" } }

⎿✓status: completed · 3 docs · 27 fields · no review required

↳workflows_versions_publish{ "workflow_id": "wf_claims_packet", "version": "draft" }

⎿✓version v12 published · production traffic 10%

●Published the claims packet workflow with policy checks and targeted review.

▎

Agents can create workflows, inspect source files, create runs, and diagnose failures autonomously.

Features

Enterprise

Run document automation with the security, privacy, and deployment controls enterprise teams need before production.

Enterprise teams scale sensitive document workflows with fine-grained access controls, SOC2 compliance, and private deployment.

Features
SOC2
HIPAA
GDPR
Audit Logs
Self-hosting

Latest

Recent product updates and releases from Retab.

BlogRetab's engineering blog

@retab/ui

Open-source UI components for your document workflows

Enterprise features

Security, privacy, and deployment controls for production document workflows.

Start building with Retab now

For teams. Launch document workflows with parsing, extraction, validation, review, and audit trails in one place.

Get a demo Read the docs

For developers. Install a Retab SDK or connect Retab MCP to your agent runtime.

CLI

Agent Skills

Retab split beats Google

End-to-end automation for your hardest document workflows.

Process documents at scale — with configurable business rules, cross-document checks, and human review for exceptions.

Start free Get a demo

For document workflows

To turn files into data

Automated by agents

Workflows

Build durable document pipelines that keep every parse, split, extraction, validation, and review step in one place.

Retab orchestrates the full document lifecycle without stitching brittle tools together.

Features
Workflow Builder
Human Review
Schema Versioning
Run Inspection

Features

Turn PDFs, scans, spreadsheets, and emails into typed data with APIs that understand layout, tables, sources, and schemas.

/parse

Acme Corp12 Market St, Boston MA

INVOICE

#2042

BILL TONorthwind Trading Co.88 Harbor Rd, Seattle WA

IssuedMar 4, 2024

DueApr 3, 2024

TermsNet 30

ITEMDESCRIPTIONAMT

A-1Widget assembly$480

A-2USB-C cable$96

A-3Mount bracket$58

Subtotal$634

Tax$57

Total$691

Authorized signature

Date

| Item | Description | Amt |

|---|---|---:|

| A-1 | Widget assembly | $480 |

| A-2 | USB-C cable | $96 |

| A-3 | Mount bracket | $58 |

| Subtotal | $634 |

| Tax | $57 |

| Total | $691 |

| Authorized signature | Date |

/extract

/edit

/split

/partition

/classify

Developers unlock mission-critical data from documents that were previously too complex to automate.

Features
Extract
Split
Edit
Parse
Classify

Reliability

Know when automation can run straight through, when it needs review, and exactly where every answer came from.

k-LLMs Consensus

Extraction Avalue 1

Extraction Bvalue 1

Extraction Cvalue 2

Likelihoods0.67%

Extraction A

{
  "loan_amount": 250000,
  "applicant_name": "Jane M. Doe",
  "debt_to_income": 0.34,
  "term_months": 360,
  "annual_income": 184000,
  "credit_score": 742,
  "property_value": 325000,
  "down_payment": 75000
}

Extraction B

{
  "loan_amount": 249500,
  "applicant_name": "Jane Marie Doe",
  "debt_to_income": 0.35,
  "term_months": 360,
  "annual_income": 181500,
  "credit_score": 739,
  "property_value": 324500,
  "down_payment": 75000
}

Extraction C

{
  "loan_amount": 250000,
  "applicant_name": "J. M. Doe",
  "debt_to_income": 0.33,
  "term_months": 359,
  "annual_income": 184000,
  "credit_score": 741,
  "property_value": 325000,
  "down_payment": 76000
}

Likelihoods

loan_amount96%

applicant_name99%

debt_to_income91%

term_months94%

annual_income93%

credit_score98%

property_value95%

down_payment97%

Evals & monitoring

0.0%

(avg)

0.00%

Accuracy

100%91%85%79%

Reviews

Review-based validation

invoice_total

0.41

Extraction result

invoice_total

$12,089.000.41

vendor_name

Acme Industrial0.97

due_date

2025-03-150.94

Operations teams approve exceptions with confidence scores, source grounding, evals, and audit-ready traces.

Features
Confidence Scoring
Source Grounding
Evals
Reviews