Configuration

Settings

Defaults for how the FDE Agent behaves in this workspace. Authorization is always per client; nothing here widens a client's scope.

Workspace
Organization
Fixed Labs
Signed in as
Sarah Chen · Forward Deployed Engineer
Agent
FDE Agent (persistent, account-scoped)
Default permission mode

Ask answers without touching systems. Plan proposes a sequence for review. Execute carries out the plan within authorized scope, still stopping at every Class C gate.

Agent behavior

  • Computer use

    Allow browser sessions on client systems that have no API.

  • Persistent memory

    Retain client context, systems and conventions between sessions.

  • Auto-promote runbooks

    Promote a procedure to a runbook after a single verified run.

  • Expanded tool detail

    Show full tool inputs and outputs in chat by default.

Permission classes
  • Class AReversible
    Autonomous
    ResearchRead systemsGenerate codeDraft documentationRun tests
  • Class BControlled write
    Autonomous within approved scope and fully logged
    Create workflowsModify stagingOpen pull requestsCreate ticketsUpdate scoped CRM records
  • Class CConsequential
    Human approval required
    Production deploymentsPermission changesCustomer-facing communicationDeleting records
  • ExcludedExcluded
    Never autonomous
    Financial transfersLegal determinationsEmployment decisionsUnrestricted credential access
About

The FDE Agent is a digital forward deployed engineer: a persistent coworker assigned to a client account. It inspects real systems, plans work, builds and tests, verifies its own output, and escalates consequential decisions to a named human owner. It works across the whole Fixed Labs client path — Assessment, Implementation, Fractional AI Team and Technical Staffing — rather than as a separate service.

Judgment stays with humans. Authority is granted per system, per client, and is always revocable.