Skip to main content

Product tour

See every step a governed agent change goes through

The real product, in the order you would build it: the models it routes through, the schemas that guard each handoff, the pipeline graph, the human gate, the agent dispatch, and the record it leaves behind. Self-hosted - the same software you run yourself.

Every screenshot on this page is a capture of a live Modulo instance.

Step 1

Connect the models you already use

Register a model backend - Anthropic, OpenAI, or any OpenAI-compatible endpoint such as Ollama or vLLM. Modulo health-checks each backend and routes around a failed provider, and individual nodes choose the backend they run on.

Model Backends admin page showing two configured providers with health status

Step 2

Define the shape of every handoff

Bind a versioned schema to each node's input and output. A step whose output does not match its schema fails before the next node runs, so a malformed agent response never reaches the tool behind it. A schema change creates a new version and leaves the old one in place.

Schema detail view showing typed fields and version history

Step 3

Compose the pipeline as a graph

Place agents, connectors, gates, and human steps as nodes on a canvas and connect them with edges. The editor validates the graph before it saves - missing bindings, incompatible schemas, and unreachable nodes are rejected there, not at run time.

Pipeline editor canvas with a multi-node graph showing agents, connectors, and gates connected by edges

Step 4

Put a human where the risk is

Mark a step human-only and the run pauses in Awaiting Human. A reviewer claims the gate, reads the run context and the upstream output, and approves or rejects. The decision, the reviewer, and the timestamp are recorded.

Step 5

Dispatch the agents you already run

Modulo does not supply agents. It dispatches work to the runtime you already use - Claude Code, Cursor, Copilot, or your own harness - and writes the prompt, output, cost, and exit status to the run record.

Completed run detail showing the run status, total cost, and captured input and output

Step 6

Prove what happened

Every governance event - run start, agent dispatch, human decision, gate outcome - lands in a hash-chained, append-only audit log inside your instance, next to the full captured output of each agent run. Verify the chain, or export it for an audit.

Audit log with tamper-evident events and Verify Chain button visible

Run it yourself

Modulo is self-hosted software. Clone it, bring your own models and agents, and keep every prompt and every record inside your own network. There is no hosted trial.