AI where the work already happens.

Use language models as ordinary pipeline steps: summarise, classify, extract fields from PDFs and images, let an agent choose tools, or search your own data. Bring your own model provider.

Start free Book a demo

Useful, not a demo

Extract invoice fields, classify tickets, draft replies — inside the same flow that files them.

Works with AI assistants

Call MCP tools from a pipeline, and expose your pipelines and data as tools for assistants like Claude.

Costs you can see

Model routing, memory and per-project budgets keep spend predictable.

What’s included

Capabilities

LLM step with JSON mode and templated prompts
Extract: text and structured fields from PDFs and images
Agent step that picks tools within set limits
Embeddings and vector search (pgvector)
MCP client and MCP server / data gateway
Describe a pipeline in plain English and review the generated draft
Anthropic, OpenAI and compatible providers

Example

Invoice inbox

1New email with a PDF invoice
2Extract supplier, totals and lines
3Match against the purchase order
4Flag differences to finance

Works well with

Visual pipelines

Drag steps onto a canvas. Branch, loop, run in parallel and merge back.

Connect everything

Databases, APIs, files, email, directories, servers — the systems you already run.

Triggers & schedules

Run on a schedule, on a webhook, on a new file or record, or on demand.

Stop moving data by hand.

Start free with three projects, or talk to us about running it across your organisation.

Start free Book a demo