FlyLab

Agent runtimechecking site tools · r1

Recommended when connectedfind_fly_circuits

Human gatenot applicable · required after design

Agent handoff

Read-only in this browserCopy the current revision, exact input references, and recovery path as one machine-readable packet.

Runtime diagnosticchecking

Document ready: not exposed
Secure context: not exposed
Origin-keyed cluster: not exposed
Tools policy: not exposed
modelContext present: not exposed
registerTool type: not checked
Registration attempted: no
Declared tools: 8
Registered now: checking
Registrations accepted: 0/8
Failed tool: none
WebMCP invocation observed: no
Page session
State revision: r1
Recommended next tool: find_fly_circuits

Registration exceptionnot applicable · API unavailable before registration

Visible recovery packet

{
  "schema_version": "flylab.agent-handoff.v1",
  "transport": {
    "schema_version": "flylab.agent-runtime.v1",
    "page_session_id": null,
    "api": "document.modelContext.registerTool",
    "status": "checking",
    "page_registration_status": "checking",
    "page_invocation_handler_available": false,
    "webmcp_invocation_observed": false,
    "webmcp_client_availability": "unavailable",
    "declared_tool_count": 8,
    "registered_tool_count": null,
    "agent_invocation_available": false,
    "workflow_next_tool": "find_fly_circuits",
    "invocable_next_tool": null,
    "invocable_next_action": {
      "callable": false,
      "blocked_by": "webmcp_availability_checking"
    },
    "fallback": {
      "mode": "read_only_same_tab_dom",
      "mutation_available": false,
      "browser_documentation_url": "/agent",
      "manifest_url": "/flylab-agent-manifest.json",
      "contract_url": "/flylab-tool-contracts.json",
      "live_context_selector": "#flylab-agent-context",
      "handoff_selector": "#flylab-agent-handoff",
      "scope": "same_open_page_only"
    },
    "execution_note": "The JSON documents and inline state are read-only discovery aids, not a fallback transport. Use a compatible WebMCP runtime to call tools.",
    "capability_diagnostic": {
      "schema_version": "flylab.webmcp-capability-diagnostic.v1",
      "document_ready_state": null,
      "secure_context": null,
      "origin_agent_cluster": null,
      "permissions_policy_tools_allowed": null,
      "document_model_context_present": null,
      "register_tool_type": null,
      "registration_attempted": false,
      "registrations_accepted_before_rollback": 0,
      "failed_tool_name": null,
      "registration_error_name": null,
      "registration_error": null,
      "availability_reason": "checking"
    }
  },
  "workflow_recommendation": {
    "tool": "find_fly_circuits",
    "kind": "tool",
    "input_refs": {},
    "reason": "Begin with the bounded source-backed circuit search."
  },
  "agent_context": {
    "schema_version": "flylab.agent-context.v3",
    "page_session_scope": "current_open_page",
    "agent_status": "ready",
    "state": {
      "revision": 1,
      "stage": "discover",
      "goal": "Investigate how the adult fruit-fly brain coordinates leg and wing output during rapid escape. Separate measured findings from connectome inference and simulation assumptions, draft a falsifiable hypothesis, and design a controlled experiment. Stop for my approval, then continue, analyze every metric, compare conditions, and save the complete evidence bundle."
    },
    "artifacts": {
      "discovery_decision_id": null,
      "selected_circuit_id": null,
      "discovered_evidence_ids": [],
      "hypothesis_eligible_evidence_ids": [],
      "causal_evidence_ids_by_perturbation": {
        "activate": [],
        "silence": []
      },
      "hypothesis_id": null,
      "hypothesis_evidence_ids": [],
      "hypothesis_predicted_behavior": null,
      "hypothesis_perturbation": null,
      "experiment_id": null,
      "experiment_approved": false,
      "approval_binding_complete": false,
      "approval_experiment_id": null,
      "approved_protocol_hash": null,
      "approved_seed_manifest_hash": null,
      "approval_timestamp": null,
      "condition_ids": [],
      "batch_id": null,
      "analysis_ids": [],
      "analysis_metrics_by_id": {},
      "comparison_id": null,
      "comparison_analysis_ids": [],
      "evidence_bundle_id": null
    },
    "artifact_manifest": {
      "model": {
        "id": "FlyLab mapped-motor embodiment model@0.3.0",
        "artifact_type": "model_manifest",
        "provenance": [
          "derived"
        ],
        "parent_ids": [],
        "evidence_ids": [
          "E-FLYLAB-MODEL-004",
          "E-GF-SEQUENCE-013"
        ],
        "source_ids": [
          "SRC-FLYLAB-MODEL-CARD",
          "SRC-FLYGYM-NM-2024",
          "SRC-FLYGYM-CODE-V210",
          "SRC-GAITANIDIS-PLOS-BIOLOGY-2025"
        ],
        "name": "FlyLab mapped-motor embodiment model",
        "version": "0.3.0",
        "controller": "state-coherent-mapped-circuit-adapter.v2",
        "environment": "stateful-open-field-model-scale.v3",
        "controller_mapping_provenance": {
          "provenance": "agent_hypothesized",
          "statement": "Each mapping from a source-backed circuit path to a body controller is hand-authored and versioned; it is not an inferred firing rate, synaptic simulation, muscle model, or optogenetic dose."
        },
        "calibration_status": "literature_constrained_event_order_unfitted_amplitudes",
        "calibration_summary": "State-transition order and approximate GF event intervals are constrained by cited adult escape measurements; probabilities, amplitudes, controller gains, recovery timing and dynamics, and MDN dynamics remain hand-authored and are not fitted to held-out data.",
        "parameterization": {
          "name": "flylab.mapped-motor-parameters.v3",
          "provenance": "agent_hypothesized",
          "calibration": "State-transition order and approximate GF event intervals are constrained by cited adult escape measurements; probabilities, amplitudes, controller gains, recovery timing and dynamics, and MDN dynamics remain hand-authored and are not fitted to held-out data.",
          "calibrationStatus": "literature_constrained_event_order_unfitted_amplitudes",
          "unitBoundary": "Distances and speeds use declared model-scale millimeter units. They are internally consistent but are not biologically calibrated effect sizes.",
          "durationReferenceMs": 1800,
          "durationGainBounds": [
            0.35,
            1.2
          ],
          "unilateralGain": 0.72,
          "maximumMotorDrive": 1.1,
          "silencingReferenceMotorDrive": 0.72,
          "maximumSuppressionFraction": 0.92,
          "reverseProbability": {
            "baseline": 0.08,
            "driveGain": 0.79,
            "minimum": 0.02,
            "maximum": 0.97
          },
          "signedSpeed": {
            "forwardBaselineMmS": 0.92,
            "forwardDrivePenaltyMmS": 0.25,
            "reverseInterceptMmS": 0.62,
            "reverseDriveGainMmS": 2.25,
            "jitterScaleMmS": 0.36,
            "forwardJitterScaleMmS": 0.28
          },
          "responseLatency": {
            "interceptMs": 540,
            "inverseDriveGainMs": 1320,
            "jitterScaleMs": 260,
            "minimumClampMs": 180
          },
          "backwardDistanceScale": {
            "minimum": 0.72,
            "maximum": 0.92
          },
          "heading": {
            "baseDeg": 11,
            "driveGainDeg": 34,
            "bilateralJitterScaleDeg": 7,
            "unilateralJitterScaleDeg": 12
          },
          "stanceStability": {
            "baseline": 0.91,
            "drivePenalty": 0.08,
            "jitterScale": 0.06,
            "minimum": 0.62,
            "maximum": 0.98
          },
          "reverseWalk": {
            "legRecruitment": {
              "baseline": 0.18,
              "driveGain": 0.72,
              "jitterScale": 0.06
            }
          },
          "escapeTakeoff": {
            "responseProbability": {
              "baseline": 0.04,
              "driveGain": 0.9,
              "minimum": 0.01,
              "maximum": 0.98
            },
            "responseLatency": {
              "interceptMs": 1.4,
              "inverseDriveGainMs": 3.2,
              "jitterScaleMs": 0.4,
              "minimumClampMs": 1.4
            },
            "eventTiming": {
              "controllerLeadMs": 0.6,
              "groundReleaseDelayMs": 1.1,
              "wingDelayAfterGroundReleaseMs": 1.5,
              "recoveryBaseMs": 180,
              "recoveryDriveGainMs": 120,
              "recoveryJitterScaleMs": 20,
              "sourceIds": [
                "SRC-GAITANIDIS-PLOS-BIOLOGY-2025"
              ],
              "boundary": "The order and approximate millisecond intervals are literature-constrained calibration targets across distinct direct-GF and light-off paradigms; they are not a fitted equivalence to FlyLab unitless drive."
            },
            "verticalDisplacement": {
              "interceptModelMm": 0.42,
              "driveGainModelMm": 2.6,
              "jitterScaleModelMm": 0.24
            },
            "wingRecruitment": {
              "baseline": 0.05,
              "driveGain": 0.9,
              "jitterScale": 0.08
            },
            "legRecruitment": {
              "baseline": 0.06,
              "driveGain": 0.92,
              "jitterScale": 0.08
            },
            "forwardSpeedGainModelMmS": 0.7,
            "trajectoryLiftGainModelMm": 0.035,
            "trajectoryForwardGainModelMm": 0.018
          },
          "trajectory": {
            "steps": 80,
            "reverseDriveThreshold": 0.12,
            "baseStepModelMm": 0.018,
            "driveStepGainModelMm": 0.036,
            "positionJitterScaleModelMm": 0.006,
            "headingJitterScaleDeg": 0.08,
            "activeTurnBaseDegPerStep": 0.32,
            "activeTurnGainDegPerStep": 0.34
          },
          "stateTrajectory": {
            "protocolWindowSemantics": "[onset_ms, min(trial_duration_ms, onset_ms + duration_ms))",
            "distanceScale": {
              "minimum": 0.98,
              "range": 0.04
            },
            "eventSamplingBoundaryEpsilonMs": 0.001,
            "reverseControllerLead": {
              "latencyFraction": 0.2,
              "maximumMs": 60
            },
            "stanceStability": {
              "preparationPenalty": 0.05,
              "reverseWalkPenalty": 0.1,
              "jump": 0.35,
              "wingDeployment": 0.25,
              "airborne": 0.5
            },
            "takeoffPose": {
              "jumpPitchDeg": -14,
              "wingDeploymentPitchDeltaDeg": 20,
              "airborneRecoveryPitchDeg": 6,
              "wingDeploymentLegDecayFraction": 0.75,
              "airborneLegRetentionFraction": 0.2,
              "unilateralBodyRollPerHeading": 0.2
            },
            "illustrativeCompatibilityPose": {
              "takeoffPitchDeg": 10,
              "airborneStanceStability": 0.6
            }
          }
        },
        "boundary": "Literature-constrained GF event order and selected approximate intervals with hand-authored, unfitted reduced-order probabilities, amplitudes, body-controller gains, and recovery timing and dynamics. It does not execute connectome neurons, electrical or chemical synapses, neural dynamics, muscles, aerodynamics, FlyGym, or a wet-lab perturbation, and it is not independent biological validation."
      },
      "discovery_decision": null,
      "selected_circuit": null,
      "discovered_evidence": [],
      "hypothesis": null,
      "experiment": null,
      "approval": null,
      "batch": null,
      "analyses": [],
      "comparison": null,
      "evidence_bundle": null
    },
    "next_action": {
      "kind": "tool",
      "name": "find_fly_circuits",
      "callable": false,
      "blocked_by": "webmcp_availability_checking",
      "reason": "Begin with the bounded source-backed circuit search.",
      "input_refs": {},
      "workflow_preconditions_satisfied": true,
      "workflow_blocked_by": null,
      "callability_scope": "current_page_transport_and_workflow_state"
    },
    "next_tool": null,
    "human_controls": {
      "next_trial_budget": 2
    },
    "human_gate": {
      "id": "approve_experiment",
      "status": "not_applicable",
      "subject_experiment_id": null,
      "blocks_tool": "run_fly_simulation",
      "agent_can_satisfy": false,
      "webmcp_tool_can_satisfy": false,
      "scope": "webmcp_site_tools",
      "authorization_boundary": "Approval is absent from the WebMCP tool surface and requires visible UI interaction. It is not identity-authenticated against general browser automation.",
      "human_control_label": "Approve experiment",
      "approval_experiment_id": null,
      "approved_protocol_hash": null,
      "approved_seed_manifest_hash": null,
      "approval_timestamp": null,
      "follow_up_execution": "not_authorized"
    },
    "pipeline": [
      {
        "name": "inspect_flylab_state",
        "title": "Inspect shared state",
        "kind": "tool",
        "status": "available",
        "boundary": "Read-only recovery of this page session, artifact IDs, blockers, and the next valid action."
      },
      {
        "name": "find_fly_circuits",
        "title": "Search source-backed evidence",
        "kind": "tool",
        "status": "recommended",
        "boundary": "Curated adult evidence only; no simulation is run."
      },
      {
        "name": "draft_fly_hypothesis",
        "title": "Draft a falsifiable hypothesis",
        "kind": "tool",
        "status": "blocked",
        "boundary": "Creates an agent_hypothesized claim, not evidence."
      },
      {
        "name": "design_stimulation_trial",
        "title": "Design controlled conditions",
        "kind": "tool",
        "status": "blocked",
        "boundary": "Creates a visible draft protocol with controls, timing, replicates, and seed."
      },
      {
        "name": "human_approval",
        "title": "Operator reviews exact protocol",
        "kind": "human_gate",
        "status": "blocked",
        "boundary": "Intentionally not a WebMCP tool. Any protocol edit revokes approval and clears downstream work."
      },
      {
        "name": "run_fly_simulation",
        "title": "Run the approved virtual batch",
        "kind": "tool",
        "status": "blocked",
        "boundary": "Reduced-order simulation only; no wet-lab action or neural dynamics."
      },
      {
        "name": "analyze_fly_behavior",
        "title": "Quantify simulated behavior",
        "kind": "tool",
        "status": "blocked",
        "boundary": "Metrics aggregate summaries derived from complete seeded state trajectories. The arena replays the exact selected run; the separately serialized condition illustration is compatibility-only and excluded from analysis."
      },
      {
        "name": "compare_fly_trials",
        "title": "Rank and propose a follow-up",
        "kind": "tool",
        "status": "blocked",
        "boundary": "May propose one bounded follow-up; execution is not authorized."
      },
      {
        "name": "save_fly_evidence",
        "title": "Save the complete lineage",
        "kind": "tool",
        "status": "blocked",
        "boundary": "Creates a provenance-rich browser-local bundle and portable JSON export."
      }
    ],
    "interpretation_policy": {
      "reported_empirical_observations": "measured",
      "deterministic_catalog_and_method_records": "derived",
      "structural_wiring": "connectome_inferred",
      "generated_trajectories": "simulation_predicted",
      "draft_protocols": "agent_hypothesized",
      "calculated_metrics": [
        "derived",
        "simulation_predicted"
      ],
      "proposed_claims_and_follow_ups": "agent_hypothesized"
    },
    "provenance_policy": {
      "definitions": {
        "measured": "An empirical result reported under the cited study conditions; it is not automatically universal or reproduced by FlyLab.",
        "derived": "A deterministic filter, aggregation, or calculation from pinned source records; it is not a new wet-lab measurement.",
        "connectome_inferred": "An anatomical inference from EM reconstruction or annotations; it does not establish activity, physiological efficacy, or behavior.",
        "simulation_predicted": "An output conditional on the stated model, controller mapping, parameters, and seed; it is not biological validation.",
        "agent_hypothesized": "An untested proposal authored by an agent that must remain distinct from source evidence and simulation output."
      },
      "inheritance": "Each artifact_manifest record labels its complete scientific subtree unless a more specific nested record provides its own provenance.",
      "operational_boundary": "State revisions, blockers, next actions, UI labels, and storage references are operational metadata rather than scientific evidence.",
      "untrusted_annotation": "Caller-supplied goals, titles, and notes are administrative input and are never counted as scientific provenance."
    },
    "session_warning": "Tool artifact IDs belong to this open page session. Inspect state again after interruption, edits, cancellation, or navigation.",
    "workflow_next_tool": "find_fly_circuits"
  },
  "trust": {
    "untrusted_human_fields": [
      "agent_context.state.goal"
    ],
    "authority": "Re-inspect this same open page after interruption, visible edits, cancellation, or navigation. Artifact IDs are page-session scoped."
  }
}

Raw machine endpoints remain /flylab-agent-manifest.json and /flylab-tool-contracts.json. Some in-app browsers block top-level JSON navigation; the HTML guide carries the same static documentation without leaving this page session.

Mission boundary

Human-owned

Published at r1. The agent should call inspect_flylab_state before acting.

Local workflow0/8

Inspect shared stateinspect_flylab_stateavailable

Search source-backed evidencefind_fly_circuitsrecommended

Draft a falsifiable hypothesisdraft_fly_hypothesisblocked

Design controlled conditionsdesign_stimulation_trialblocked

Operator reviews exact protocolhuman_approvalblocked

Run the approved virtual batchrun_fly_simulationblocked

Quantify simulated behavioranalyze_fly_behaviorblocked

Rank and propose a follow-upcompare_fly_trialsblocked

Save the complete lineagesave_fly_evidenceblocked

Activity1

Shared audit activity

checking site tools

Laboratory readysystem · r1

Waiting for a behavior goal or a WebMCP tool call.

Manual recovery walkthrough

fallback using the same validated actions. This uses the same local action handlers and validation rules. It does not exercise Site Tool discovery, registration, or agent invocation.

Local lab workflow 0/8WebMCP transport 0/8 · not attempted · checking

Optional visual audit

Open-field trial · Three.js 3D fly

posteriorright5 model mm

Awaiting protocolpreview

Three.js WebGL schematic external morphology

Loading 3D fly…

Design and approve an experiment to generate seeded trajectories.

0 ms

5,000 ms

no protocolno model target windowawaiting design

Experiment queue

No protocol yet

No condition artifacts exist yet. A valid design must add baseline and model-sham controls plus the requested perturbation arm; bilateral designs add left-only and right-only comparisons only when the selected motor map supports them.

Exact protocolNot created

Exact protocol for operator review

Not created

No protocol artifact exists. The design tool must create a hypothesis-linked, controlled protocol before approval becomes available.

Bounded autoresearchNext-experiment replicate budget · 2

Bounded autoresearch

propose only

Next-experiment replicate budget

The agent may rank and propose a follow-up. It cannot execute a new batch without approval.

Artifact stater1

Agent runtime contract

checking

Workflow recommendation onlyfind_fly_circuitsNot callable in this browser. Use a compatible WebMCP runtime, then inspect fresh state.

State: r1

Evidence boundaries13 records

Evidence boundaries

MMeasuredDDerivedCConnectome inferredSSimulation predictedAAgent hypothesis

Literature-constrained GF event order and selected approximate intervals with hand-authored, unfitted reduced-order probabilities, amplitudes, body-controller gains, and recovery timing and dynamics. It does not execute connectome neurons, electrical or chemical synapses, neural dynamics, muscles, aerodynamics, FlyGym, or a wet-lab perturbation, and it is not independent biological validation.

FlyLab agent workspace

Open-field trial · Three.js 3D fly

No protocol yet