Keep the objective intact.
A request becomes a plan with context, constraints and a definition of done — not a prompt handed to an isolated model.
The AI operating system that learns you and evolves itself→Beyond answers.
Beyond answers.
Into action.
Say what you need, and TheOne gets it done — code, research, documents, scheduled runs, long-running goals. It asks before it acts, and shows the proof when it's done.
Start in your browser. Bring the desktop app when the work needs your files. · Desktop 0.4.37 for macOS and Windows
- 10
- capabilities
- 93
- built-in skills
- 22
- automation templates
- 6
- delivery channels
- 3
- coding engines
ObjectiveWhat matters
MemoryWhat persists
PolicyWhat is allowed
EvidenceWhat is proven
THEONE / WORKSPACEPRODUCT PREVIEW
Bind a workspace, describe a change, and review the work.
Start with a question, a document, or an objective.
Recurring work that runs on schedule and asks before acting.
Preferences, facts and decisions every conversation uses — yours to edit.
Product screenshots with example content. Not a live task or a performance benchmark.
ROOT INTELLIGENCE
The layer above execution.
Agents are powerful at doing one kind of work. TheOne keeps the larger objective, context and rules in view, then coordinates the work across them.
OBJECTIVEMEMORYPOLICYTOOLS
ROOT INTELLIGENCEFrame the work
before agents actone source of intent
before agents actone source of intent
CodexcodingClaude CodecodingOneClawtools
Route work to the right agent.
Coding, research and tools can use different specialists while the workspace, approvals and history stay connected.
Return a result with a record.
The run shows what happened, what changed and which checks passed. You can review the evidence instead of trusting a black box.
ONE CONTEXT. MANY AGENTS.
Different specialists.
One place to steer them.
Coding, voice, research and long-running goals can share the same workspace without giving up visibility or control.
CODING
From a repository to a reviewable change.
Explore the code. Plan the change. Run tools and checks through OneClaw. Review the diff before deciding what ships.
Explore→Edit→Test→Review
Explore TheOne Coding VOICE
Say it naturally.
Speak, follow the captions, and handle approvals on screen. Voice is another way into the same workspace.
Meet TheOne VoiceGOALS & AUTOMATION
Give the work a next step.
Keep a goal moving after you close the tab, or schedule a recurring run. Results go where your team already works, with measures, a budget and a review record.
Explore goals ↗RESEARCH & DOCUMENTS
Turn context into something useful.
Investigate a question, work through documents, and create an artifact you can review and use. Keep sources and assumptions visible.
Explore research ↗THE TOOLBOX
The tools can vary.
The context stays.
Explore the capabilities that agents can bring into a governed run — from coding and research to memory, integrations and long-running goals.
Coding
Changes your code, and proves it works
Learn more →Voice
Talk it through, hands free
Learn more →Research and writing
Reads widely, cites its sources, writes it up
Learn more →Documents and design
Real files and real designs, not just text
Learn more →Goals and automations
Keeps working after you close the tab
Learn more →Memory
Remembers what matters, and shows you what it remembers
Learn more →Skill library
93 built-in skills, picked for the task
Learn more →Integrations and API
Results where you already work, and a gateway for your own agents
Learn more →Markets lab
Research, risk and paper trading — never real orders
Learn more →Self-evolution
Gets better from its own failures, under rules it cannot change
Learn more →TASK ORCHESTRATION
One objective. The right agent. A verifiable result.
TheOne turns delegation into a visible sequence — so you can see where a task is, what it needs and what remains to be checked.
- 01
Understand
Turn the request into an objective, constraints and a definition of done.
- 02
Authorize
Match the work to the right specialist while keeping the workspace and history connected.
- 03
Execute
Choose the workspace, review the plan and approve gated actions before they run.
- 04
Review
Inspect the output, evidence and unresolved risks instead of trusting a final message.
GOVERNANCE PROOF
Autonomy you can audit.
Evidence stays with the work.
Powerful tools need clear boundaries. TheOne separates proposing an action, executing it and verifying its outcome — then keeps the record available for review.
Read about governance ↗You set the boundaries.
Workspaces, approvals and budget controls keep the scope of a task explicit.
Evidence, not a victory message.
Review diffs and checks. Execution success and independent verification are different states.
Improvement without self-approval.
Verified failures can inform improvement proposals. Evaluation and release approval remain separate gates.
LONG-TERM MEMORY
Keep the context.
Keep the choice.
Preferences, facts, decisions and lessons can carry from one conversation to the next. Search, edit, pin or forget them whenever you choose.
See how memory works →SELF-EVOLUTION
It learns from its failures.
It cannot rewrite its rules.
Verified failures become improvement proposals. Each change replays the same real tasks as the current version and ships only if nothing regresses. Its permissions and verifier are never self-modified.
How self-evolution works →
WEB AND DESKTOP
Use it where you work.
One account across both. Start in the browser; add the desktop app when you want TheOne to work on the folders on your own computer.
TheOne on the web
Chat, voice, research, writing, goals and automations, from any browser. Nothing to install.
TheOne Desktop
Everything on the web, plus coding directly in the folders and repositories on your Mac or PC.
Version 0.4.37 · macOS (Apple silicon) and WindowsSTART SMALL. SEE THE VALUE.
Your first useful task.
No elaborate setup needed to explore. Copy a starting point, open TheOne, and bring your context.
Understand a codebase
Review this repository in read-only mode. Explain its structure, identify one important issue, and propose a small fix before changing any files.
Research a question
Compare three approaches to this project. Separate sourced facts from assumptions and finish with a practical recommendation.
Plan an objective
Help me turn this objective into a plan with clear milestones, an explicit budget, and checkpoints that need my approval.
Suggested starting points. Nothing runs on this website.
What should I try first?
Start with a small, read-only task: review a codebase, summarize a document, or draft a plan. Review the output before granting broader access.
Do I need the desktop app?
You can start in the browser. The desktop app adds access to authorized folders and repositories on your own computer. Check Downloads for supported builds and installation notes.
Can I choose the coding engine?
The architecture supports the native agent and configured Codex or Claude-based engines. Actual availability depends on the connected runtime, credentials and your plan.
Is every answer independently verified?
No. Verification depends on the workflow and its configured checks. Inspect the run's evidence and verification status; a completed response alone is not a verified result.
How much does it cost?
See the live pricing page for current plans, model access and usage limits. This website does not duplicate prices that could become outdated.
ONE INTELLIGENCE. EVERY AGENT.
Give the next task a governed path.
Web and desktop. One account. Built by OneAI Labs.