Glossary

This is the master glossary for askTheodor — every term you’ll meet in these docs, from the AI and networking jargon to askTheodor’s own ideas, in one alphabetized list. Each entry is a single plain sentence; follow the screen and concept pages for the full story behind any of them.

Looking for one word?

Press ⌘K (or Ctrl+K) to open search and jump straight to a term, or use your browser’s find (⌘F / Ctrl+F) on this page.

A

• Agent — an AI worker that can reason and act by calling tools (not just chat), pursuing a goal across multiple steps.
• API key — a secret string that authenticates you to an AI provider so the app can use that provider’s models on your behalf; kept in askTheodor’s local database, encrypted at rest.
• Approval gate — a checkpoint where the app pauses and waits for your explicit OK before doing something sensitive or irreversible (like sending an email); auto-denies when running headless.
• Autonomy — a worker’s ability to act unattended (run Plans and Routines on its own), always bounded by budgets, limits, the approval gate, and the kill-switch.

B

• Budget — a spending ceiling (global or per-Company) that caps how much money workers may consume; a hard-stop blocks further runs once it’s hit.

C

• Cache — a stored result from an identical earlier tool call, reused to avoid repeating work and spending tokens again (marked with a 🔁).
• Channel — a connection to an outside messaging surface (email, WhatsApp, etc.) that workers can read from and draft replies into.
• Command Center — a full desktop install of askTheodor used by one operator; the cockpit you steer from and the only kind of instance that sends outreach or makes final approvals.
• Company — a named grouping of workers, with its own Library and settings, that organizes your workforce into teams.

D

• Daemon — a long-running background process with no user interface; an askTheodor Team-Hub runs as one on a server.
• DAG — a directed acyclic graph: a way of wiring a Plan’s tasks so each one only starts after the steps it depends on, with no circular loops.
• Delegate — to hand a task or chat to another paired instance to run, with the result reported back.

E

• Embedding — a numeric representation of a piece of text’s meaning that powers semantic search and RAG retrieval in the Library.

F

• Fleet — the screen and system for seeing and controlling all your paired instances — their status, sync, and delegation.

G

• GGUF — a file format for packaging a local AI model so a runner like Ollama or the built-in engine can load and run it on your own hardware.

H

• Hard-stop — an automatic halt that blocks a run from continuing, most commonly when a budget ceiling is reached.
• Heartbeat — the periodic status ping an instance sends so the Fleet screen knows it’s alive and what it’s doing.
• HMAC — Hash-based Message Authentication Code: a cryptographic signature on each request between instances that proves it came from a paired peer and wasn’t tampered with.

K

• Kafenio — askTheodor’s group chat space (a “café”) where you and multiple workers can talk together in one room.
• Kiosk — a locked-down, single-purpose mode of askTheodor (a paid add-on) for exposing one narrow interface, separate from the full app.

L

• Library — your curated knowledge base of documents, notes, and links; the source of the material workers quote via RAG.
• Local-first — the principle that your data lives on devices you own (a local database) rather than on a vendor’s cloud.

M

• MCP — the Model Context Protocol, an open standard for connecting tools and data sources to AI; askTheodor is both an MCP client and server.
• Memory — what a worker retains across conversations — its own notes and facts — so it doesn’t start from scratch each time.

O

• Ollama — a popular program for running open AI models locally on your machine, usable as a provider in askTheodor.
• Orchestrator — the role that coordinates other workers, drafts Plans, and reports back; in askTheodor this is Theodor.

P

• Peer — another askTheodor instance yours is paired with, able to sync and exchange delegated work over signed requests.
• Persona — a configured AI character (a Worker): a name, role, instructions, allowed tools, and memory.
• Plan — an orchestrated, multi-step workflow — a set of tasks assigned to workers and run in order by the scheduler — that you approve once.
• Provider — the company or program that supplies the actual AI model (OpenAI, Anthropic, Ollama, the built-in engine, …); you connect at least one.

R

• RAG — Retrieval-Augmented Generation: the app fetches relevant Library items and feeds them to a worker so its answer is grounded in your material rather than invented.
• Reflection — a worker pausing to review its own progress or output and adjust, improving quality on longer tasks.
• Routine — a scheduled or triggered job a worker runs on its own (e.g. “digest my inbox every morning”).
• Runner — headless mode (asktheodor --runner): the app with no window, suitable for running a Team-Hub on a server.

S

• Scope — the level of access something has: a pairing’s capability grant (sync / delegate / keys / full) between instances, or which workers a Library item is shared with.
• Sync — the background mirroring of the shared workspace (Companies, Workers, Library, Plans) across paired instances.

T

• Team-Hub — a headless --runner instance (usually on a server) that runs your workhorse workers 24/7; the paid “remote” add-on.
• Theodor — the built-in orchestrator (“chief-of-staff”) worker that coordinates the others, drafts Plans, and reports to you.
• Token — the unit AI models read and write in (roughly ¾ of a word); providers bill per token, which is why the app tracks them.
• Tool — a capability a worker can invoke beyond generating text (web search, file read/write, browser, make a document, and more).
• Tribe Doctrine — askTheodor’s guiding philosophy that you run a workforce of specialized, cooperating workers under your command, not a single chatbot.
• Tunnel — a service (e.g. cloudflared or ngrok) that gives a machine a reachable inbound URL so peers can connect to it through firewalls or NAT.

W

• Worker — a configured AI persona that does real work for you; many workers, organized into Companies, make your workforce.

Terms in this page

This page is the glossary — every term above is its own definition. For the longer explanations, see the screen and concept pages each term links out from.