FixControl/Documentation

Admin Guide

Infrastructure & system

How FixControl reaches your code (SSH and the desktop bridge), the retention windows you control, and what your FixControl operator controls rather than you.

This page covers the plumbing: how FixControl reaches code that is not on a git host, which retention windows are yours to set, and which knobs sit with your FixControl operator rather than with you.

Infrastructure connections

Settings → Infrastructure is where you connect code that does not live on GitHub, GitLab or Bitbucket. The subtitle states the boundary: "Connect a credential and pick which repositories your agents work on. Your agents can only access codebases activated below."

Two connection kinds live here:

  • SSH — a server FixControl reaches over SSH to read a repository or folder.
  • Local (desktop bridge) — code on a machine that is not reachable from the network, brought into reach by the FixControl desktop app. The screen says Desktop app required and offers the download when you pick it; it is not a hidden prerequisite.

GitHub, GitLab and Bitbucket are not connected here. They live under Settings → Channels → Code & patch delivery, and the Infrastructure page links there — see Git providers. Other product connections (Slack, Microsoft Teams, mailboxes and the rest) are under Settings → Channels too; see Integrations.

Once a connection exists, link the repository to a project with Project → Connections → Connect repository, or register it under Codebases.

Private Kubernetes, Jenkins, GitLab and Argo are a different mechanism again: nothing is connected from here. Under the agent connection model none of them is exposed to FixControl — that is the outbound-only FC Agent.

Workspace choices

A handful of choices your workspace makes for itself — among them the language for messages sent outside the app — live on Settings → Overview & progress under Settings for this workspace. See Switching workspaces.

Retention you control

Your organization's retention windows live in the governance console: Settings → AI & approvals, tab Audit & Retention (shown with Show advanced settings). Each window is in days, and 0 means keep indefinitely — the default for all of them.

WindowWhat happens when it runs outShortest window you can set
Agent tracesDeleted7 days
ApprovalsThe free text and evidence attached to approval decisions are redacted; the decision itself stays90 days
Runtime logsDeleted7 days
Customer communicationMessage bodies are redacted; metadata and hashes stay30 days

Audit events are not on that list. The audit-event window is enforced by FixControl for the whole platform: 365 days on the hosted service, and never below 90 days. The console shows that field as Platform-managed; a value entered for your organization there is advisory and does not shorten or lengthen what is kept. The reason is structural: the audit hash chain is a single global sequence, so it is pruned in one place, and pruning records an anchor so the entries that remain stay verifiable across the gap.

What retention never removes: approval decisions, orchestration events, and the audit chain anchors. Deleting the evidence of a decision is not a retention policy, and the database refuses it — see Governance & policy.

What your FixControl operator controls

Some settings are not in your workspace at all. They live in the FixControl operator's own console, which is visible only to FixControl staff, and they cover the platform underneath every workspace:

  • System housekeeping — the retention pass over finished agent runs and the audit window, the audit chain check that goes with it, and when the sweep runs.
  • Issue key sequences — the prefixes new issue keys are drawn from, and which issue kinds get their own sequence. If you want a different prefix, ask your FixControl operator to set it up.
  • AI providers and models — which providers and models FixControl uses, and the credentials it holds for them. You do not configure AI provider accounts or API keys. What an organization admin can choose is a preferred model per agent, on AI Team → the agent → Orchestration settings (see Orchestration & context).
  • Platform behaviour — sandbox execution and its resource caps, queue backends, search weights and retrieval depth, discovery and preview budgets, the per-preview hostname template, embedding providers, whether the FC Agent and remote validation are available to your workspace, and the database used for verification runs by default.

Three facts about that console are worth knowing from your side:

  • A value has a visible source. Each setting shows whether it is the platform default, a value set centrally, or one inherited from the deployment's own configuration. There is no invisible override.
  • Some settings are per workspace. Where a setting supports it, the operator can set a different value for your workspace without changing anyone else's.
  • Some knobs are deliberately not settings. Anything that would weaken container isolation, relax host allowlists or disable fail-closed database isolation is kept out of that console on purpose — a form is the wrong place for "switch the safety off", and those stay part of how the deployment itself is installed.

If you need one of these changed, ask your FixControl operator.

FAQ

Why can my agent not see this repository? Either no connection reaches it, or it is not linked to the project you are working in. The two are separate on purpose — see Codebases.

Does retention delete my audit trail? Audit events are kept for the platform-wide window — 365 days on the hosted service, never below 90 — and pruning records an anchor so the remaining chain still verifies. Nothing on your side shortens it.

Can I keep agent traces forever? Yes. Leave Agent traces at 0, which is the default.

Related: Codebases · Git providers · Integrations · FC Agent · Governance & policy · Security & audit

Anything unclear or wrong?Let us know →

FixControl is a trade name of FixControl B.V.