Crawling...
Reactive Integration Platform
Self-composing
connectors
Matcha is a reactive runtime for connectors. Connect your systems, write typed state, and everything downstream updates automatically — durable, traceable, and decoupled.
Traditional Workflows
A
push
B
push
C
Node A must know about B, and B about C. Tightly coupled, fragile chains.
vs
Matcha Document Triggers
A
reacts
B
reacts
C
A just writes state. B reacts to changes. C reacts to B. Zero coupling.
The Core Idea
Workflows that behave like spreadsheets
In a spreadsheet, you don't route values between cells. You declare dependencies, and updates propagate automatically.
Matcha applies that same principle to your infrastructure:
- Connections write state
- Document Triggers declare what state they react to
- State changes create durable execution signals
- Outputs become new state, which triggers the next document triggers
- The graph is generated from contracts, not hand-wired
No orchestration code. No message routing. No coupling between steps.
How It Works
Three primitives, infinite systems
01
Connect
Plug in Gmail, REST or Swagger APIs, webhooks, and web crawlers. Each connector imports external data as typed state with IDs and versioned updates.
02
React
Document triggers listen for state changes and evaluate before scheduling work. If conditions match, execution is scheduled durably — guaranteed at-least-once across crashes and restarts.
03
Propagate
Workers execute asynchronously at their own pace. Results are written as new or updated state, which triggers downstream document triggers automatically. Ingestion stays fast, processing scales with your capacity.
Built for Orchestrated AI
Agents become document triggers
Matcha makes AI a first-class part of your connector runtime.
Run an agent inside a document trigger to classify, extract, plan, or validate. Agents subscribe to state types, emit new state, and return evaluations that trigger the next steps. Run multiple agents in parallel, compare results, and route low-confidence outputs to human review — all with full traceability.
Gmail
Emails arrive. Attachments get parsed. Data flows.
Connect a mailbox and Matcha imports emails on a schedule. Messages and attachments are normalized and classified. Purchase data, receipt totals, contract terms — all become typed state that downstream document triggers can process.
Inbox
EML + Attachments
Parsed State
Inbox
Acme Corp
Invoice Q4-2024
Supplier AB
PO #4521 confirmed
Nordic Ltd
Contract renewal
PDF
invoice_q4.pdf
VendorAcme Corp
Amount12,450 SEK
Date2024-01-15
REST / Swagger APIs
Point at a spec. Matcha discovers and syncs.
Give Matcha a Swagger endpoint and it discovers every entity — customers, orders, products, line items. Each entity becomes typed state with versioned updates. Sub-entities and binary fields are fetched and stored automatically. The API becomes a live data source.
Swagger Spec
Entity Discovery
Typed State
swagger.json
GET/api/customers
GET/api/contracts
POST/api/orders
GET/api/products
Customers
Contracts
Products
Documents and Data
Drop any document. Get structured data back.
Matcha keeps imported files, records, and integration payloads connected to typed document state. AI-assisted classification and workflow rules route data to the right downstream triggers.
Raw File
OCR + AI Parse
Fields + Tables
Webhooks & Crawlers
Events push in. Pages get scraped. Document triggers fire.
Receive real-time webhook payloads from any system — each becomes typed state that triggers document triggers immediately. Or crawl websites on a schedule: detect content changes, import new pages, and react to updates automatically.
External Event
State Write
Document Trigger Chain
{ "event": "order.created" }
{ "event": "payment.settled" }
{ "event": "user.signup" }
example.com
Why Matcha
Built for production, not demos
Full Traceability
Every change and every execution is visible end to end, from input to final output.
Resilient by Default
Crashes, retries, and backpressure are normal. Matcha is designed for them.
Zero Coupling
Producers never reference consumers. Document triggers depend on state types, not upstream steps. Add, remove, or upgrade processing without breaking flows.
Self-Hosted
Run on your infrastructure. Air-gapped capable. GDPR friendly. No cloud dependency, no vendor lock-in.
AI-Powered Classification
Schema-aware classification, mapping, and workflow routing — built in.