The platform

How aXentic actually works.

Studio builds and runs your agents. Shield governs every action they take. Both run on one runtime.

Studio

Build and run agents on any model.

  1. 01 Agents

    Any model, swappable

    Bound to your knowledge and tools, on any major LLM — switch providers without a rewrite.

  2. 02 Workflows

    Durable workflows

    On Temporal: fourteen step types, human waits included, and a run survives a restart.

  3. 03 Xens

    Digital workers

    Persona-driven agents with lasting memory — chat on Slack and Telegram, and join live calls.

  4. 04 Knowledge

    Hybrid retrieval

    Keyword and vector search, scoped per agent and per document.

  5. 05 Tools & connectors

    80+ tools, plus MCP

    Built-in search, evidence, approvals and memory — each tool and connector scoped to the agents allowed to use it.

  6. 06 Platform Assistant

    aXentic builds aXentic

    Describe what you need, and it drafts the agent or workflow for you to review.

Shield · govern and observe

Three axes. One decision on every action.

Every action passes all threePolicy says may. Autonomy says how far. Obligation says must. And every check lands in the work graph.

Shield · Shadow-AI discovery

Five places we look for the AI nobody registered.

Network egress

Calls to LLM APIs, traced to a source.

Cloud access

Accounts holding LLM permissions.

Code pipelines

Repos importing LLM SDKs.

Container images

Images with LLM SDKs inside.

Billing and cost

LLM spend nobody approved.

CorrelatedRisk-scoredBrought under governance

Shield · Data governance

Inside the run, not at the network edge.

01 Retention-aware routing

Your data never trains someone else’s model

Restricted data bound for a provider that would keep it is held for approval.

02 Mid-run re-gate

It can stop a leak mid-thought

Data more sensitive than the run was cleared for is re-judged before it lands.

03 Classification inheritance

Sensitivity that spreads on its own

Anything derived from a Restricted record inherits its tier.

04 Governed retrieval

The knowledge lookup can carry a verdict

Each retrieved fact carries its sensitivity, checked before it leaves.

One runtime

Governance in every stage, not just at the edges.

BuildDeployRunPost-runSHIELD · GOVERNS EVERY STAGEONE AUDIT TRAILOwner required to approveGate + autonomy ceilingPolicy before, halt mid-runEvidence, not logs
Most platforms govern at two moments — before the build, and after the run. aXentic governs inside all four stages.

One runtime · The three graphs

Not a log — three graphs, one enforcement spine.

Graph 1 · Audit

The work graph

How was this produced?

Every run, decision and approval, linked by what produced what.

Graph 2 · Govern

The object registry

What touched this record?

The business objects agents act on — classified, validated and traced.

Graph 3 · Reason

The knowledge graph

What’s connected to what?

Connected facts your agents reason over, each carrying its sensitivity.

One classification-and-lineage spine under all three

Bring us an agent — or an idea.

Book a working session →Apply as a design partner