Sovereign AI Workflow Platform

Controlled AI workflows for document-heavy operations and finance, with sources, policy checks, approvals, owners, and evidence on European infrastructure.

Choose the audience that matches your question

The page contains three selectable readings of the same subject.

For consumers

Fabric brings your questions, files, notes, links, mail, and calendar context into one private workspace. Ask in everyday language, inspect the sources behind an answer, and turn the work you repeat into a routine you can keep using.

For businesses

Dweve turns document-heavy work into a controlled flow from intake to approval. Sources, policy checks, exceptions, owners, actions, and evidence stay in one record, so operations can clear blockers and finance can measure the gain before scaling.

For engineers

A request begins as a durable Fabric object and invokes only the product boundaries the work needs. Agent, cognition, knowledge, policy, target, execution, and placement events return with the result, each tied to a named contract.

    return Err(SessionExpired);

One request · explicit handoffs · evidence returned

The route changes with the work. The identity, contracts, and return path stay explicit.

Only the layers a request needs have to run. This view shows the complete path so every boundary can be inspected.

result + evidence return to the work surface

The products do not hide what they depend on

Only relationships asserted by the current product and open-source pages are drawn here. Absence from this ledger is not a claim that a foundation is unused.

Inside the product or a real dependency of it.

A supported integration, not an internal dependency.

Foundations with no product edge asserted in this view

Numerus, Signum, and Selvedge remain part of the 14 open-source routes. This ledger does not invent a product relationship where the current pages do not make one.

The route may shorten. The thread must not break.

Analyzed request, identified 3 key concepts

Retrieved context from the Dweve Loom knowledge base

F(n) rounds phi to the n over root five.

Ratios of terms approach phi, near 1.618.

The sequence appears across nature and tiling.

Iterative growth, linear in n, exact in integers.

Dweve Loom, Query Analysis, 3 domain specialists, 2 sources

Type a message or @ to mention an agent...

Fabric brings your questions, files, notes, links, mail, and calendar context into one private workspace. Ask in everyday language, inspect the sources behind an answer, and turn the work you repeat into a routine you can keep using.

One workspace for your context, answers, and recurring work.

Dweve turns document-heavy work into a controlled flow from intake to approval. Sources, policy checks, exceptions, owners, actions, and evidence stay in one record, so operations can clear blockers and finance can measure the gain before scaling.

Baseline one workflow, run real work, and compare the result.

A request begins as a durable Fabric object and invokes only the product boundaries the work needs. Agent, cognition, knowledge, policy, target, execution, and placement events return with the result, each tied to a named contract.

Inspect the representations, constraints, computation, placement, and claim boundaries instead of trusting a model name.

Start at the trace, the responsibility map, or a computation contract.

Pre-order Fabric Personal for EUR 14.25 or Fabric Friends & Family for EUR 35.63 for five named people. The pre-order price remains fixed for the uninterrupted contract. Access opens by group; it is not public yet.

Choose real work with a known baseline, real documents, a named owner, and an approval. Compare cycle time, queue age, repeat work, exceptions, and evidence effort before deciding how far to expand.

Follow a request trace, inspect the system map, or test a handoff at the knowledge, policy, execution, or placement boundary. Judge the architecture by the state and evidence it returns.

Your personal Fabric interface. Chat with memory, upload files, build small assistants, and save recurring routines through the public Dweve Mesh.

No. Your conversations and files stay yours. We do not train models on your content, ever.

In the EU only. The Business Workspace uses EU-only paths on the public Dweve Mesh, so your data is not routed through other jurisdictions.

Fabric Personal is EUR 19 a month for one named person. Fabric Friends & Family is EUR 47.50 for five named people, the price of 2.5 Personal subscriptions. Both use Fabric through the web and app without Aura, direct API access or Agent SDK rights. The compute-contribution route is limited to eligible Personal customers after the first paid month.

Yes. Every answer shows where it came from. You can open the sources behind any response instead of taking it on faith.

Bring your mail, your calendar, your files, and the apps you use every day into one place, so Fabric can work with what is already there instead of starting from a black box.

Personal is EUR 19 a month and Friends & Family is EUR 47.50 for five named people. Their pre-order prices are EUR 14.25 and EUR 35.63. Both are billed by a Dutch entity and cancelable at any time. After the first paid month, an eligible individual Personal customer may instead use the compute-contribution route; Friends & Family is excluded.

Sovereign AI with clear deployment options.

European infrastructure under European law.

Private, on-prem, or air-gapped environments.

After the first paid month, eligible customers may use Fabric's published spare-compute route under its current terms.

EU jurisdiction, controlled deployment, no sovereignty asterisks

The review packet forms while the work moves. Source, rule, owner, and replay stay attached to the decision, so a later audit starts with evidence instead of reconstruction.

Can this decision still be defended after the people and systems around it have changed?

Yes. The result carries the source used, the rule applied, the named reviewer, and a replayable record of the path taken.

Arithmetic, hashing, representation, and reasoning each keep an explicit contract. The result is reproducible because every layer preserves the same invariants, not because a final API says it is.

A discrete decision can name the facts it accepted, the rule that fired, and the branch it rejected. Inspect the mechanism first; only then judge the product claim built on top.

Three premises satisfied the policy rule; one conflicting branch was rejected; the accepted path is attached to the result.

Normalised facts enter as explicit values

The policy boundary is evaluated directly

Conflicting branches are recorded, not hidden

The accepted route leaves a portable receipt

The premises satisfy the named policy rule

The conflicting branch fails the boundary check

A single speedup hides where the work went and what trade-off was made. This run profile keeps workload, unit, hardware, and guard test beside each figure, so the shape stays visible and comparable.

Eight commercial surfaces do not turn every request into an eight-product route. Fabric keeps the work record coherent while the operating case decides which cognition, knowledge, agent, execution, placement, and accountability boundaries participate. Kera target portfolios remain separately agreed.

Fabric holds the record; Charter keeps rights and responsibilities explicit.

Loom composes cognition; Spindle governs versioned knowledge.

Nexus coordinates multi-agent work; Aura is the software change agent.

Core executes supported workloads; Mesh places eligible work. Kera is scoped separately.

The work record is shared. Product participation remains conditional.

The useful audit trail is the one assembled during the decision. Sources, exceptions, accountable owners, approvals, and returned actions stay attached to the same record, so review begins with evidence rather than reconstruction.

The version and relevant passage stay beside the claim.

Open risk stays visible until an owner resolves or accepts it.

The accountable person and approval state remain named.

The action and its evidence can be inspected or exported together.

Review the record that produced the decision, not a summary written later.

The production open catalogue contains fourteen foundations with different responsibilities. A shared language or repository is not the contract. The input shape, interface, invariant, and conformance result are the contract; Jacquard remains adjacent research.

Types, grammar, premises, events, or model state are explicit.

The API, ABI, or portable artifact defines the handoff.

A replacement preserves behaviour, ordering, or proof scope.

Pinned cases test substitution at the boundary.

Replace an implementation only after the declared boundary still passes.

Reasoning, execution, system provenance, and agent activity are related but not interchangeable. Keeping their scopes separate is what makes an end-to-end trace inspectable without pretending that one receipt proves everything.

Records reasoning steps and checkable proof against supplied premises.

Records what code ran, from input through sealed host calls to output.

Records append-only events across the wider system.

Records signed agent events and folds them into a replayable run.

Compose the records for a wider trace. Do not substitute one scope for another.

Graph-native language and toolchain when selected.

Separate product · not a ninth suite surface

Jacquard, Forge, and Mycelia sit beside the production route. Their presence here does not assert a runtime dependency.

A separate graph-native systems product, selected for its own contract.

Separate product · outside the eight-product suite

Jacquard, Forge, and Mycelia are research tracks beside the portfolio. This board does not claim that the suite depends on them.

Jacquard, Forge, and Mycelia are shown together because all three belong to the research track. No runtime relationship to the fourteen foundations is asserted here.

All connectors funnel through the same audit + permission layer.