Platform · Open source · Auditable
Ch∆In
AI workflows you run yourself — locally, air-gapped or distributed, provable step by step, up and running in under 60 seconds.
Ch∆In is available. Selected partners already receive access to the sources; the source code is released as open source in Q3 2026.
What Ch∆In is
You have to use AI — but you cannot give up control of your data, and you have to prove every step. That is exactly what Ch∆In is built for.
AI tasks run as traceable workflows on your own infrastructure — from a single machine to a federated hub mesh across sites. Everything that happens is recorded — complete and tamper-evident — in the audit log.
And it is fast and reactive: Ch∆In handles not only documents but also audio, video and live streams.
Up and running in under 60 seconds
One command, one binary — on Linux, macOS and Windows. From nothing to your first run in under a minute.
Also via Homebrew or containers — or skip the terminal entirely: chain studio opens the visual wire editor.
How it works
Build a flow
Wire steps together by drag-and-drop in the visual editor — or write them as YAML. Both stay in sync.
Run it
Locally, air-gapped or as a federated hub mesh (mTLS). Every step isolated in the WASM sandbox, deterministic and reproducible.
Prove it
A complete, tamper-evident audit log — with the flow hash, the model digest and every approval.
Modules & store
Ch∆In gets work done through modules — ready-made building blocks anyone can find, install and inspect in the Ch∆In store.
Extract text
PDF and DOCX become searchable text.
Translate
Multilingual — locally or via a model.
Summarise
Long documents down to the essentials.
LLM chat
Ask a language model, locally or via API.
Approval
A human decides before it continues.
Integrations
MCP, n8n and more — connected as modules.
The Ch∆In store: every module signed and versioned. Build your own with the Rust SDK — one function is enough; the WASM Component Model keeps the door open to more languages.
Bridges bring whole ecosystems in as modules: 400+ n8n integrations, 100+ MCP servers and SPARK sources — each becomes an auditable module the moment it is registered.
Fits into your stack
What you already use stays usable: every n8n, MCP or SPARK source brings its workflows in as modules automatically — plus many of our own.
n8n
Trigger workflows by webhook from a flow.
MCP
Model Context Protocol tools (e.g. Claude Desktop) as steps.
SPARK / Temporal
Call Temporal-backed, long-running workflows directly as a step.
HTTP / REST
Connect any API — method, headers, auth.
What teams use it for
License audit
Read an SBOM, classify licences, get approval, generate a report — for procurement and OSPO.
Meeting anonymisation
GDPR-compliant: produce a redaction plan, have it reviewed, apply it.
Document processing
Extract, a human reviews, then summarise — with the model digest in the log.
Security & compliance
- ✓Tamper-evident audit log via hash chain
- ✓WASM sandbox: no network or filesystem without permission
- ✓Every release with SBOM and signature — CRA-ready
- ✓Human-in-the-loop approvals with attribution
- ✓Runs locally, air-gapped or federated
- ✓Apache-2.0 — fully auditable
Visuals coming soon. 
Visuals coming soon. 
Visuals coming soon. 
Visuals coming soon. 
Visuals coming soon. 
Visuals coming soon. Sources & access
The source code is released as open source in Q3 2026. Until then, selected partners receive early access to sources and documentation.
Request early access →