Wud / control plane + leaf agents

Give every workflow a leaf.

Wud coordinates portable browser agents that research, use tools, and carry out multi-step work. Configure what a leaf can do, give it a task, and keep the run in view.

Portable HTML + JavaScript leavesInspectable runs, tools, and resultsControlled permissions and approvals
WUD / COORDINATE
leafcarry out work
memoryretain context
runrecord progress
QORTEX / REASONING
E2B / ISOLATED TOOLS
01 / WUD

Coordinate

Configuration, schedules, work records, identity, and backend access.

02 / LEAF

Carry out tasks

A browser workspace a person can use to research, act, and review results.

03 / QORTEX

Power reasoning

AI inference and agent capabilities reached through the control plane.

A browser agent with a home base

The leaf works. Wud keeps the system together.

Each leaf is a build-free HTML and JavaScript agent. Wud supplies the shared services around it, so a task has a place to start, a set of permitted capabilities, and a durable record when the work is done.

  1. 01

    Shape a leaf

    Set its purpose, reasoning tier, tools, skills, and operating limits.

  2. 02

    Give it work

    Open its workspace and describe a task, or invoke a defined routine. Bring focused GitHub access to the named project and issue work.

  3. 03

    Let it act

    The leaf reasons through Qortex and calls only the capabilities it is allowed to use.

  4. 04

    Inspect the result

    Follow events, tool activity, artifacts, usage, and the recorded outcome. Review run progress, approvals, files, and results. If work is interrupted, use its recorded progress to decide how to continue.

What a leaf can do

Useful work, with the edges visible.

Give each leaf the tools its job needs. Wud validates inputs, records activity, and keeps consequential actions reviewable.

01 / WEB

Research the live web

Use ranked web search and current web-grounded answers to investigate a question or gather source material.

web.search · web.current
02 / FILES

Work in an isolated sandbox

Read, write, list, and execute commands in an E2B environment when a task needs files or shell work.

sandbox.read · sandbox.exec
03 / MEMORY

Keep durable notes

Store and recall agent-scoped memory, then retrieve reusable skills that shape how a leaf approaches a task. Reuse project context and instructions — keep saved information and skills connected to ongoing agent work, so important instructions and project details are close at hand.

memory.remember · skills.get
04 / CONNECTIONS

Connect approved services

Use fixed, permissioned adapters such as GitHub without turning a public connection token into a provider secret. Bring focused GitHub access to the job — connect GitHub for project and issue tasks within the access granted to the agent.

github.issues.list · github.issues.create
05 / OUTPUT

Leave an inspectable trail

Runs retain events, checkpoints, tool invocations, artifacts, and usage so outcomes are traceable and reviewable.

run.events · run.checkpoints

The control plane

Turn one useful task into a system you can reuse.

Save a routine with the tools it needs, then connect it to a schedule or an event. Reuse that starting point while Wud keeps ownership, permissions, run progress, and failures visible.

Help shape the early release
WUD / WORK OVERVIEW
LEAVESconfigure
RUNSinspect
SCHEDULESrepeat
routine / weekly research briefqueued
leaf / approved tool activitytracked
artifact / source summaryretained

Portable by design

The agent belongs where the work happens.

Wud is designed around a portable leaf bundle: HTML, JavaScript, styles, and public connection metadata. Host it at a URL, register that deployment, and keep provider credentials on the backend.

YOUR HOSTLeaf webpagePerson-facing workspace
WUDControl planeIdentity · runs · schedules
CONNECTED SERVICESQortex + E2BReasoning · isolated tools

Early access note: interactive leaves are the current center of gravity. Reliable unattended execution across independently hosted leaves is being completed through the worker, wake-up, and deployment path.

A fit for work with a shape

Start with the job you repeat.

RESEARCH

A brief that returns with sources

Give a leaf a research objective, let it gather live material, and keep the resulting work attached to a run.

OPERATIONS

A routine your team can inspect

Define the steps, parameters, and permitted tools once. Reuse the routine without hiding the execution trail.

MAINTENANCE

A connected task with a boundary

Pair approved service connections with explicit grants and review points before a leaf changes durable state.

Questions before the trunk opens

Common questions

What is a leaf?

A leaf is the HTML and JavaScript agent a person actually uses. It runs the task loop in a browser and reports progress and results through Wud.

Is Wud the agent?

Wud is the control plane. It coordinates configuration, identity, tools, runs, memories, schedules, and backend services; the leaf carries out the person-facing work.

Can I host a leaf myself?

The product is designed around downloadable, build-free leaf bundles and registered deployments. Early access will help harden ownership, origin verification, and operational health across hosted leaves.

What happens when a browser sleeps?

Interactive work depends on an available leaf browser. Scheduled and event-driven work needs a supervised worker and explicit unattended authorization; that production path is still being completed.

How does billing work?

Wud can use the shared Forgewise wallet and QreditCharge flow when configured. Early access will help complete reconciliation and production billing coverage.

Can someone receive access to a shared agent while the owner retains ownership?

Give a specific person access to your agent while you retain ownership and control of its rules. Decide separately which connected tools that agent may use.

Can I require approval for selected actions?

Yes. Require approval for selected actions, with decisions attached to the specific request and the person authorizing it. Keep meaningful choices visible as work moves forward.

Can I set limits on how long a task runs and how many actions it takes?

Yes. Set limits for how long a task runs, how many actions the agent takes, and how many tasks run at once. Keep agent work within the boundaries you choose.

Early access

Put a leaf on the work.

Tell us what you would want a browser agent to handle. We'll use the list to shape the first useful release.

No launch promises or noisy marketing. Just early-access updates about Wud.