Skip to content

Features

humanize runs flows: directories of Python that drive one or more coding agents in a loop and write down everything they did. Most backends drive a coding agent you already have under its existing login; the bundled DeepSeek Harness is the SDK-backed exception.

This section is what the system is rather than how to operate it: the unusual parts drawn, one map over all of them, and a page for each mechanism worth pushing on. Nothing here is a command — the guide that runs it is one click from every diagram.

To install it and run something, start on the home page, which has a quickstart for running a flow, weaving one and working on humanize.

A run, as it happens

One flow, many agents, one trace. Every turn's tool calls land on the timeline as they are made — every agent, every sub-agent and every program those turns ran, on one clock. Hover a lane; change how many agents are on it.

runninghmz exec -f flame_chase
claude/claude-opus-5:highwrites the kernel0 tok
codex/gpt-5.6-sol:highreads what landed0 tok
dsh/deepseek-v4-pro:highruns the benchmark0 tok
kimi/kimi-code/k3:highports the module0 tok

4 agents · 0 slices on screen · one track per row of an agent's sessions

A simulation of the shape of a run, not a recording. One timeline is how the real one is built.

What it does, one picture each

The agent runs here. Its syscalls land there.

A seccomp-filtered ptrace supervisor decides every call the coding agent makes, one at a time. No plugin, no configuration, no cooperation — the agent is told none of it.

this machinethe targetclaude · codex · dsh · …unchanged, and told none of thissyscallssupervisorseccomp filter · ptraceevery call decided one at a timeanswered herecredentials · state · the model providerssh · docker · tcp · a pipehmz internal anchor servefilescontents, renames, modes — the target's own errorsprocesseseverything the agent spawns, in the target's cwdthe networkwhatever those commands reach

lands on the target Read out of the local mirror at local speed. The mirror and the target are kept in step.

How it works, syscall by syscall: The anchor. What you are deliberately not entitled to: its reference.

How the capabilities fit together

A run crosses five systems: the flow that describes the work, the control plane that drives agents, the fabric that decides where work lands, the record that keeps it continuous and readable, and the surfaces people start and inspect it through. Hover a group to read the guarantee it owns.

AFlow systemexpress · prove · compose · resume
A1Expression & compilationstart hereFree-form Python shares one runtime with an atlas, whose restricted body compiles into a graph.A2Static correctness & provingStructural mistakes are rejected before a model call costs time, money, or a live run.A3Composition & hot reloadFlows and skills can nest; regular flows can reload current source between calls.A4Scheduling, state & resumptionPlacement and fan-out pair with explicit state or atlas node-level resumption.
BAgent control planesessions · tools · recovery · identity
B1Backend unificationDifferent CLIs and app servers expose one session protocol without flattening their capabilities.B2Turn & session controlstart hereTyped, capability-aware controls steer, clone, and pursue goals where a backend supports them.B3Tools & skillsSkills are selected per session, while flow callbacks can become temporary native tools.B4Failure recoveryA failed session can recover in place, migrate accounts, or use another CLI without losing intent.B5Accounts & credentialsCredential inputs can be isolated, redirected, and reused without leaking provider state.
CExecution fabriclocal control · remote work
C1Transparent remote executionstart hereA local agent operates a remote machine within documented process and signal boundaries.C2Shadow workspace & consistent writesRemote workspaces appear immediately and writes land atomically as files arrive on demand.C3Portable transport runtimeTargets need no install: one protocol carries processes, files, environment, and working directory.C4Machine lifecycleMachines can be isolated per agent or shared for a run, with explicit lifecycle ownership.
DRun continuity & observabilitydetach · recover · reconstruct · scrub
D1Detached operationstart hereRuns outlive terminals: a workspace daemon preserves PTYs, replay, attach, and stop control.D2Persistent state & layered logsState and nested journals are written through, so a crash leaves a readable recovery record.D3Trace reconstructionSessions, sub-agents, and processes rebuild onto one calibrated, session-bounded timeline.D4Telemetry privacyReporting has explicit consent state and is scrubbed before anything leaves the machine.
EProduct surfacesdiscover · configure · invoke
E1Discovery, forking & configFlows can be discovered locally, forked atomically, and configured from their schemas.E2Unified entry pointsstart hereSDK, CLI, terminal interface, and daemon reach the same underlying flow and run model.

5 systems, 19 capability groups. Hover or focus a group for its core guarantee; open it for the closest explanation.

The complete map adds the boundaries, related guides and reference for every group: Capability map.

Feature deep dives

Each takes one mechanism far enough that its trade-offs make sense, around a diagram you can push.

Flow system

Python becomes a prophecyA deliberately narrow flow becomes a typed graph that can be checked, compared and resumed node by node.
A flow is PythonOrdinary Python and compiled atlases live side by side, chosen by how much of the work must be knowable before it runs.
Many turns at onceTurns are sequential inside one session; concurrency comes from having several conversations to run.
Picked up where it stoppedOrdinary flows preserve explicit state; atlases preserve completed node visits. Neither recreates a conversation.

Agent control plane

Many backends, one agentNative servers, streaming CLIs and Agent Client Protocol backends meet one session contract.
Two accounts of one CLICredentials, model catalogues and failure chains stay isolated while a session changes where it runs.
A line typed mid-turnAcknowledged queues put guidance into the turn that is working rather than behind it.
A turn can be cut offA per-turn cap on tokens or clock, read off the live meter, ends the turn that is running.
Every run has an allowanceHours, millions of output tokens and dollars, held to at every session edge, stop the whole run.
Answers in a shapeA pydantic model is both the question and the contract the answer must satisfy.
It decides when it is doneA backend-owned pursuit loop continues until the model settles the objective.
The moments of a turnTyped lifecycle moments let a flow react without teaching the backend about the flow.
You, as one of the agentsQuestions, the mission board and a person-shaped agent put human decisions on the same run.

Execution fabric

The anchorA local agent can work against a remote target while paths, processes, networks and ownership keep their meaning.

Run continuity and observability

The terminal can leaveA workspace daemon owns the PTY, so watchers may disconnect and return without owning the run.
One timelineAgent events, sub-agents and sampled processes are reconstructed on one calibrated clock.

Product surfaces

One system, four ways inLocal discovery, schema-driven setup, Python, CLI, TUI and the daemon all reach the same run and session model.

The flows it comes with

A ralph loop and a stateful one, two agents alternating, an actor with a reviewer between its rounds, a loop the model itself decides is over, and isolated lanes under a coordinator. Each has a page with its own loop played on it.

Every one of them, with the shape of each: Flows. What a weaver writes to add another: Writing a flow.

Where to go next

Before you point one at a repository you care about

humanize runs every agent with permission prompts disabled, and nothing turns them back on. A flow is trusted Python: loading or running it may execute its code, even though static checks can inspect selected structure without doing so. Read Security.

Released under the Apache-2.0 licence.