An agentcan makethe decision.
Submit a proposal. See the possibilities. Inspect the decision. Follow what happens next. TAHTO gives autonomous agents persistent identities and visible decision histories.
- tahto · Finnish
- will
- Sequence
- Propose → Decide → Authorize → Execute
Abstract idle state. No decisions are fabricated — when a genuine decision source is connected, real records replay here.
- § 01
A proposition enters
A human or application submits a proposal — an action for an agent to consider. It arrives as a single, discrete input.
- § 02
Possibilities emerge
The agent's policy expands the proposal into the outcomes it could commit to. Every branch is a candidate the agent genuinely weighs.
- § 03
One route is selected
The configured mechanism evaluates the candidates and commits to exactly one. The committed route is drawn in cobalt; the rest remain visible as unrealized possibilities.
- § 04
The boundary resolves
The committed decision crosses a formal authorization boundary. Authorization is the agent's recorded permission for the action — a distinct, inspectable state.
- § 05
Execution is separate
Authorization is not execution. Unless genuine execution infrastructure is connected, the sequence terminates truthfully at AUTHORIZED — no action is claimed to have run.
Terminates at authorized
What did the agent decide — and why?
TAHTO is an authorization layer, not a chatbot. It treats a decision as a first-class, permanent object: something proposed, weighed, committed, and — where infrastructure allows — carried out. Each stage is visible and distinct.
- Propose
- An action is submitted to an agent. The proposal is the discrete unit of intent — structured according to the agent's capabilities.
- Decide
- The agent evaluates the proposal against its genuinely configured policy. The mechanism — rule, model, or hybrid — is named accurately and left inspectable.
- Authorize
- A decision resolves into an authorization state: proposed, rejected, or authorized. Authorization is recorded permission — nothing more is implied.
- Execute
- Only when real execution infrastructure exists does an authorized action proceed. Execution is a separate state, never conflated with authorization.
Four surfaces onto the same decision record.
Every surface reads from one truthful source. When no records exist, each shows an intentional empty state — never invented activity.
Decision ledger
A chronological record of every genuine decision — proposal, agent, authorization, and execution state, in an institutional ledger.
Agent directory
Persistent identities. Each agent exposes its policy, capabilities, decision history, and — where genuine — its wallet.
Propose an action
Select an agent, describe the action, provide structured parameters, review, and submit. Nothing is signed without explanation.
The architecture
How TAHTO models agents, policies, decisions, authorization, and execution — and why each remains a distinct, inspectable state.
The environment is live. The decision record reflects only what is genuinely connected.
- Agent registry
- 0
- Recorded decisions
- 0
- Evaluation backend
- not connected
- Execution infrastructure
- not connected
No agent registry or decision log is connected to this environment yet. No agents, decisions, or executions are fabricated to fill the interface — absent data is shown as an explicit unavailable state.