Guides

Task-oriented walkthroughs for everything you'd want to do in Cerevisor.

Pick the guide that matches what you're trying to do. Each one is self-contained; you don't need to read them in order.

Building workflows

The bread and butter. How to design, configure, and run multi-agent workflows.

Add and configure agents: every field in the Agent Config popup, when to fill which, smart defaults.
Connect agents: handoffs, auto-connect, manual wiring, handoff formats.
Attach markdown files: input files, output files, the markdown panel, the drop zones.
Control flow: conditions, loops, gates: the four connection flow types and when to use which.
Run and monitor: the Run button, the Execution Modal, pausing, stopping, replaying.
The board: the notes agents leave each other during a run, asking one agent a question, and your own seat on the board.
Use the chat builder: natural-language workflow construction, attaching skills + files, the floating window.
Progressive Mode: describe what you want in chat and Cerevisor builds and runs it step by step — you never press Run.

Skills are reusable, focused capabilities you assign to agents. Cerevisor scans your skill folders and merges them into the Skills panel.

What are skills?: the concept, the file format, where they live.
Browse and assign skills: the Skills panel, drag-to-agent, the Skill Viewer.
Create a skill in the Workshop: templates, the editor, validation.
Skill compatibility with providers: why some skills only work with some providers.

Cerevisor is provider-agnostic. Mix Anthropic, Gemini, Claude Agent SDK, local Ollama, OpenAI Codex CLI, and Cursor in one workflow if you want.

Provider overview: the provider types, the library, defaults, per-agent overrides.
Anthropic: API key, model catalog, cost.
Gemini: API key from Google AI Studio, 2.5/3.x families, >200K-tier pricing.
Claude Agent SDK: Claude Code's agent engine, running inside Cerevisor.
OpenAI-compatible & local models: Ollama, OpenRouter, vLLM, LM Studio, Together, Groq, Fireworks, OpenAI itself.
OpenAI Codex CLI: device auth, sandbox modes, ChatGPT subscription.
Cursor Agent: API key, local + cloud execution.
Per-agent provider overrides: mixing providers in one workflow.
Managing the Provider Library: add, remove, set default, deletion guards.
Keeping your models current: refresh a provider's models so new ones appear in every picker, update Codex, Claude Code and Grok, and what the red spending-limit warning means.

Cerevisor remembers you across sessions: your preferences, your patterns, what's worked.

Memory overview: the four memory stores, how they interact, what's local-only.
Your user profile: editing, the current-focus block, what's auto-managed.
Harness memory & meta-cognition: what Cerevisor learns from runs, observations, self-portrait.
Disagreements & proposals: when Cerevisor wants to change your profile and you push back.
Data inventory & consent: what Cerevisor knows about you, classification by data type, the local export bundle.
Exporting your data: the right-to-portability JSON bundle.

What Cerevisor knows vs. assumes, what's changed since you last looked, what's still unknown.

Situational Awareness overview: the three views (the task, the agents, the working relationship), state labels, freshness.
Configure and run: the SA Directive, the per-agent uncertainty badge, the wave situation report, fixation warnings.

Multiple workflows (World View): workspaces, the rail, cross-workflow links, the world canvas.
Organizations (hierarchy mode): a resident team of agents with charters and durable memory — chat with any of them, watch them check with colleagues, let them work through Powers.
Scheduling & pipelines: cron schedules, multi-workflow pipelines, budget caps.

Pair and use the mobile companion: follow runs, answer prompts, build and launch workflows, view outputs, and connect from away over your own VPN.
Mobile companion security: what crosses the LAN, how pairing and revocation work, and the authority the phone never receives.

Power-user surfaces. Skip until you need them.

MCP servers: connect any Model Context Protocol server and expose its tools to your agents.
Operator: an assistant that works quietly in the background while the app is open, within a budget you set.
Running workflows from scripts: start a saved workflow from a terminal, a script or another program on this computer, read its result and answer its questions.
Factory mode: queue development items for one project and let a workflow work through them one at a time, attended or unattended, with one local commit per item.
Loop & Enhance: automatic detect-fix-rerun cycles.
Cursor cloud runs: run workflows on Cursor's cloud VMs.
Embedded apps (Windows): put Outlook, Excel, and other native apps on the canvas.
Web search providers: configure Tavily, Perplexity, Linkup.

Plans & licensing: the trial and the three plans (Free, Pro, Founding member); what each includes; how to upgrade or change plan; deactivating a seat.

The help assistant: the built-in chat that answers from these docs, your worlds and workflows, and your memory — on every plan.