Giga Dictionary
Definitions for the core terms used across Giga.
Use this page as a quick reference for Giga’s product language.
The definitions below follow the current Giga product model. Exact workspace state, permissions, provider capabilities, and model availability still come from the live environment.
Agent
An Agent is a named reusable AI role in Giga.
An Agent can have a role, instructions, context, access to permitted Connections, and Skills.
Use an Agent when a reusable operating identity should be available across multiple tasks, people, or repeated interactions.
Chat
A Chat is the direct working surface between a user and Giga.
Use a Chat to ask questions, reason through work, create or edit supported work product, use Connections, and ask Giga to take supported actions.
A Chat can be linked to at most 1 explicitly selected Track.
Chat history
Chat history is the transcript of the interaction inside a Chat.
It is separate from durable Track context such as Context Notes, Observations, Current Track State, and files.
Company Brain
Company Brain is Giga’s permission-aware company context layer.
Depending on relevance and access, it can draw from workspace and Group guidance, Tracks, Context Notes, Observations, Current Track State, files, Connections, Agents, Skills, Workflows, and user-level context.
Company Brain is structured around permissions and the work it belongs to.
Connection
A Connection gives Giga permissioned access to an external system.
Connections can represent apps, APIs, databases, MCP servers, browser sessions, messaging systems, and other supported services.
A Connection can provide live information and support actions when the current Giga and provider permissions allow it.
Context Note
A Context Note is the durable standing context for a Track.
Use it for stable background such as:
- what the Track is about
- standing instructions
- terminology
- durable constraints
- long-lived handoff context
Fast-changing progress belongs in Track memory or Current Track State.
Credential
A credential is secure authentication material behind some Connections.
Examples can include authentication tokens, API credentials, connection strings, or login material depending on the Connection method.
Credentials should stay in the secure connection flow rather than ordinary Chats, Context Notes, or files.
Learn about Managing Connections →
Current Track State
Current Track State is the curated current state and decision record for a Track.
It can capture:
- where the work stands now
- how it got there
- what was ruled out and why
- what remains open
It is mutable and should stay current as the work develops.
Database
A database is a structured store for records that are repetitive, tabular, updated field by field, filtered, sorted, aggregated, or expected to grow.
Current Giga doctrine supports SQLite and hosted Giga databases where the relevant capability is available.
Use files for authored narrative work such as reports and drafts.
File
A file is authored work product.
Files can hold reports, drafts, notes, code, webpages, exports, and other supported artifacts.
Files are separate from Track memory. A Track can have rich context and history even when its file area is empty.
Group
A Group is a placement and permission scope in Giga.
Groups can contain or scope resources such as Tracks, Agents, files, credentials, Connections, and shared context.
Use a Group when there is a real team boundary, shared area, or top-level access scope.
Memory
Memory is the durable Track context that preserves meaningful developments and helps future work understand what happened.
The Track memory model includes:
- Observations, which preserve append-only developments
- Current Track State, which keeps the curated current state and decision record
Context Notes and files serve different jobs and remain separate.
Model flexibility
Model flexibility is Giga’s product area for working across AI models and providers while keeping Giga’s workspace context and permissions conceptually separate from any single provider.
Exact providers, model availability, plan access, custom endpoints, and bring-your-own-model options must be verified from the current product configuration.
Observation
An Observation is an append-only development in Giga’s memory layer.
Observations can capture meaningful facts, changes, decisions, progress, preferences, or learned context.
They should only be attached to a Track or Group when that attachment has been explicitly selected or confirmed.
Private Group
A private Group is a Group scoped to its owner.
Each member has an existing private Group for private work.
A separate Group is generally unnecessary simply to make a Track private.
Learn about Personal vs shared work →
Pulse
A Pulse is recurring scheduled work.
Use a Pulse when Giga should run a check or task on a schedule instead of waiting for someone to ask in a Chat.
A Pulse can target an Agent, target a Track, or run without a target.
Shared Group
A shared Group is a Group used for team work.
Membership in the shared Group helps determine which people can access the resources inside its permission scope.
Skill
A Skill is reusable instructions and supporting context for a specific kind of task.
Use a Skill when the reusable asset is primarily the method, rules, standard, template, or reference context for doing the task.
Surface
A surface is a user-facing interface where someone can work with Giga.
Current official Giga surfaces are:
- Giga Web
- Slack
- ChatGPT
- Claude
- Microsoft Teams
The underlying Giga context and permissions stay consistent, while exact UI, file behavior, and available capabilities can differ by surface.
Learn about using Giga anywhere →
Track
A Track is Giga’s core long-running work primitive.
Use a Track when context, decisions, files, and progress should accumulate around an effort over time.
A Track can contain durable context across Context Notes, Observations, Current Track State, linked Chats, and files.
Track visibility currently follows the Group it lives in.
Workflow
A Workflow is a reusable multi-step process.
Use a Workflow when several meaningful ordered stages should be saved, standardized, or run again later.
Current Workflows can have a name, description, ordered steps, plain-language instructions, and optional runtime inputs.
Workspace
A workspace is the top-level Giga environment for a company or team.
Workspace membership, Groups, context, Connections, and other resources all sit within that environment.
The current workspace state is the source of truth for what a user can access.
Quick reference
| Term | Short definition |
|---|---|
| Chat | Direct work with Giga |
| Group | Placement and permission scope |
| Track | Ongoing work with continuity |
| Context Note | Durable standing Track background |
| Observation | Append-only development |
| Current Track State | Curated current state and decision record |
| Agent | Named reusable AI role |
| Skill | Reusable task method |
| Workflow | Reusable ordered process |
| Pulse | Recurring scheduled work |
| Connection | Permissioned external-system access |
| File | Authored work product |
| Database | Structured growing records |
| Company Brain | Permission-aware company context layer |