FixControl/Documentation

User Guide

Getting started

A five-minute tour of FixControl — what you see, what you can do, and where to begin.

What FixControl is

FixControl is an operator control plane for inbound work. Customer reports come in over email, Slack, or Microsoft Teams. The intake agent classifies each one. You get a per-item dashboard with structured context, suggested next steps, and — when relevant — a draft reply or a draft patch.

You stay in control. By default, nothing is sent, applied, or merged without an explicit operator action. Opt-in automation exists — policy-scoped auto-replies and per-project auto-delivery of approved patches — but it is off by default, and when enabled every automatic action is written to the audit log.

What FixControl is not

  • Not an autonomous AI agent. AI prepares the work; you approve.
  • Not a generic ticketing system. The shape of an issue and the actions per kind are opinionated.
  • Not a generic chatbot. Slack and Teams are governed action surfaces: they can answer ticket questions, route intake, collect approvals, and update cards, while every write/action follows project scope, permissions, capability health, outbox/retry, and audit. Channel messages also land as inbox rows so nothing is lost.

End-to-end flow

Channels                Intake agent           Operator review
  │                          │                        │
  ▼                          ▼                        ▼
Inbox row  ──────────►  classified  ─────────►  convert / reply
                                                       │
                                                       ▼
                                              Issue detail
                                                       │
                                                       ▼
                                          Approve / Dry run / Run fix
                                                       │
                                                       ▼
                                          Plan approval   ◄── you approve the plan
                                                       │
                                                       ▼
                                              Patch (PR or commit)
                                                       │
                                                       ▼
                                          Patch approval  ◄── you approve the change
                                                       │
                                                       ▼
                                              Apply

You see every step. You approve every meaningful gate — for risky work, the plan first, then the patch.

Three kinds of inbound work

KindWhat it isWhat you do
SUPPORTA question, request for help, "how does X work"Send a reply (you may use the AI draft, edited or as-is)
BUGSomething broken at the customerReproduce, fix, deploy — with the patch pipeline
FEATUREA request for new functionalityTriage, scope, prioritise, then build

The detail layout changes per kind. SUPPORT puts the draft reply on top. BUG puts Risks & Impact on top. FEATURE puts triage actions before any dev work.

Five-step tour

  1. Inbox. Every new inbound item. Use j/k to navigate, Enter to open.
  2. Convert to issue. Click when the row is actual work.
  3. Issue detail. Structured context: rationale, risks, sources, action items, draft reply (for SUPPORT).
  4. Action items. Concrete follow-up steps from the workflow agent. Different per kind.
  5. Send or execute. For SUPPORT, click Send in the Auto-reply panel. For BUG, click Approve or Run fix to start the patch pipeline.

Key concepts

  • Plan before patch. For risky work the AI first drafts an execution plan that you approve — you steer before there's any code, not after. Only then comes the patch, with its own approval.
  • Confidence ≠ Risk. Confidence is the model's self-rating; risk is an independent judgement on consequences. High confidence + high risk is a normal combination — read both.
  • Suggested → Auto-drafted → Auto-sent. Three escalation levels. You start at suggested. Auto-* is opt-in per surface.
  • Retry lineage. Patches that need revision link to their predecessor. After three revisions the chain pauses for human triage.
  • Dry run. Always available next to Run fix. Runs the pipeline up to apply, without committing.
  • Saved views. Per-user filter+sort+search bookmarks on every operator page.

Where to start

Anything unclear or wrong?Let us know →

FixControl is a trade name of FixControl B.V. i.o.