V
Verified Mission ControlFuture Ability · Native WebMCP
Detecting native WebMCP...

WEBMCP-NATIVE DECISION CONTROL PLANE FOR AGENTS

AI explores the options.
Humans decide the trade-offs.

Give an agent a goal, not a brittle script. Mission Control lets the agent generate and repair plans, but keeps human judgment around decisions that change cost, time, specification, or irreversible commitment. Switch domains below: the same engine and the same WebMCP tools run both.

AI finds many ways→Mission Control verifies trade-offs→Human sees only real decisions→AI repairs inside authority→Evidence proves what happened
Native WebMCPdocument.modelContext9 generic toolssame surface across domainsGeneric policy engineconstraints → decision frontierBounded authorityhuman-owned trade-offsPlan repairrecover without resetDurable provenancehash-chained + replayable
RUN THE SAME ENGINE AS:
Public repo ↗

PROCUREMENT DEMO Fulfill 500 exact-model sensors under quantity, deadline, budget, and authority constraints.

LIVE DECISION PATHReady — press Run the decision demo
1Goal
2Explore
3Plans
4Decision
5Repair
6Commit
7Evidence
NATIVE WEBMCP EXECUTION TRACESee the actual typed tool path, not just the UI story.
Checking WebMCP runtime…
Run the demo to see document.modelContext.executeTool() calls and machine-readable results.
01 · HUMAN GOAL

Goal Contract

Human-owned

The agent may search and compose plans. It cannot silently rewrite these hard constraints.

Need500 sensors
Deadline≤ 7 days
Base budget≤ $8,500
SpecificationExact model
AUTHORITY

AI may split orders and repair execution only inside a human-authorized decision envelope. Final irreversible commitment always requires explicit approval.

Contract unlocked
02 · AI EXPLORATION

Option Space

WebMCP

The agent explores the active domain.

No options explored yet

The agent must discover real constraints before it can plan.

03 · PLAN GENERATION + VERIFICATION

There is no zero-trade-off answer.

The agent generates combinations, removes impossible or dominated routes, then compresses the meaningful choices into a small human-relevant frontier.

0raw combinations→0rejected / dominated→0human-relevant plans
04 · HUMAN DECISION PACKAGE

Which constraint are you willing to trade?

The agent does not ask you to inspect every supplier. It presents only the choices that require human judgment.

Original Goal Contract is currently infeasible.

Choose one bounded exception. The original Goal Contract remains recorded; your decision creates a temporary authority envelope.

No human trade-off decision recorded.
05 · EXECUTION + PLAN REPAIR

Repair, don't restart

Bounded autonomy
WAIT
Waiting for an authorized plan

Execution starts only after the human decision package is resolved.

06 · IRREVERSIBLE ACTION

Final Authority Gate

Enforced
NOT REACHEDcommit_plan()
WAITINGFINAL_COMMIT_APPROVAL_REQUIRED

Even an authorized repaired plan cannot commit an irreversible action until the human gives final approval.

07 · DECISION PROVENANCE

Evidence Receipt

Proof
0%

No verified outcome yet.

08 · DURABLE + REPLAYABLE EVIDENCE

Decision Provenance survives the page.

Each mission is persisted in this browser as an append-only SHA-256 hash chain. Verify integrity after refresh, or replay the evidence read-only without re-executing the irreversible action.

No durable runs yet
Replay is read-only. It never calls commit_plan() again.
NATIVE WEBMCP TOOL SURFACE

Nine typed tools expose the decision loop directly to agents.

Human decisions remain human UI actions — deliberately not callable as agent tools.

WHY THIS NEEDS WEBMCP

Typed actions are the substrate; Mission Control governs the decision space.

PLAIN UI AUTOMATION

Follow the page

  • Click whatever looks like the next action
  • Hard to distinguish goal constraints from page text
  • Recovery often means restarting a flow
  • Human approval can degrade into a visual convention
WEBMCP + VERIFIED MISSION CONTROL

Operate inside a governed decision space

  • Typed tools expose exploration, planning, repair, and commit
  • Goal Contract stays machine-readable and immutable
  • Human trade-offs create explicit bounded authority
  • Evidence records why the final plan was allowed

WebMCP gives agents tools.

Verified Mission Control decides what the agent may decide.

ExploreFind many ways to reach the goal.
CompressSurface only meaningful trade-offs.
AuthorizeTurn human judgment into bounded authority.
RepairAdapt execution without silently changing the goal.