Settings
The Settings page controls workspace-wide preferences and integrations. Its tabs stay in the URL, so a link to a specific tab opens that tab directly.
| Tab | What it controls |
|---|---|
| General | Software version, concurrent scan capacity, language, timezone, time format, and the Hugging Face token |
| MCP Servers | The Classifyre MCP endpoint, bearer tokens, and client connection guides |
| Chat Bots | Telegram and Slack bots, permissions, and connection diagnostics |
| Export & Import | Portable workspace archives and transfer history |
General
General settings cover the workspace itself rather than AI behavior.
Software and scanning
The version card shows the installed Classifyre release and whether an update is available. Concurrent scans controls how many scan jobs this workspace may run at once. Higher values can sweep a large corpus faster, but consume more CPU and memory; lower values leave more capacity for the rest of the instance.
Language, timezone, and time format
These settings support Automatic browser detection or a fixed workspace-wide value.
| Setting | Options | Automatic behavior |
|---|---|---|
| Language | Automatic, English, German | Uses the browser language |
| Timezone | Automatic or an IANA timezone | Uses the browser timezone |
| Time format | Automatic, 12-hour, 24-hour | Uses the browser hour-cycle preference |
A user can override the displayed language from the header without changing the workspace default. See Instance Settings for the full resolution rules.
Hugging Face token
A Hugging Face token authenticates model downloads for audio transcription and Hugging Face custom detectors. It may come from the deployment or be saved on the General tab. Deployment-level credentials take precedence and cannot be replaced from the UI.
See Instance Settings for configuration and security details.
MCP Servers
The Classifyre Model Context Protocol (MCP) server exposes workspace data and operations to clients such as Claude, Cursor, and Codex CLI over a standard HTTP API. The tab controls the endpoint and its individual bearer tokens.
See MCP Server for client setup and troubleshooting.
Chat Bots
Chat Bots connect Telegram or Slack to Classifyre. Each bot can be enabled independently and restricted to selected capability groups and Autopilot agents. Mutation access can also be disabled to make a bot read-only. Built-in tests and diagnostics show whether credentials and the live connection work.
Export & Import
Export & Import moves selected workspace data through a compressed .cfyre
archive. Choose the data areas to include, monitor the job, and download the
completed archive; an import lets you preview the archive and select which
parts to load.
Credentials and secrets are never exported. Imported records receive new IDs, and links between those records are rewritten to remain consistent.
AI configuration
AI providers are configured in Harness AI → Configuration, not on the Settings page. That is where you assign providers independently to the interactive Assistant and the autonomous Harness, create or edit provider credentials, and run provider tests.
See AI Providers for the complete setup guide.
Harness agents
When each autonomous agent runs — immediately on new data, batched, or only once the corpus has settled — is configured per agent in Harness → Agents, alongside its system prompt and tools. Instance-wide harness limits (run budgets, evidence thresholds, and how much context each run sees) are under Harness → Config.
See Harness Agents for what each mode and gate means, and for what to check when an agent is not running.