Agents
Learn how named AI Agents in Giga combine role, instructions, context, Skills, and Connections.
An Agent is a named, reusable AI role in Giga.
Agents are useful when you want Giga to operate with a consistent role, instructions, context, and access across repeated work.
Video placeholder
Agents overview video.
What an Agent can bring together
Role and instructions
Define how the Agent should approach its work and what it is responsible for.
Context
Give the Agent durable background relevant to the work it handles.
Connections
Give the Agent access to external tools and systems when permissions allow it.
Skills
Reuse task-specific instructions and methods with the Agent.
The exact capabilities available to an Agent depend on the current workspace, Group, Connections, Skills, and live tool surface.
When to use an Agent
Create or use an Agent when you want a reusable AI role that people can return to repeatedly.
Examples include:
- a sales research Agent
- a recruiting Agent
- a customer support Agent
- an operations Agent
- a finance reporting Agent
A normal Chat works well for ad hoc work that does not need a persistent named role.
Agent vs Chat vs Skill vs Workflow
| Primitive | Best fit |
|---|---|
| Chat | Direct, ad hoc work with Giga |
| Agent | A named reusable AI role |
| Skill | Reusable instructions for a specific kind of task |
| Workflow | A reusable ordered multi-step process |
| Pulse | Recurring scheduled work |
These primitives can work together.
For example, a recruiting Agent might use a candidate-screening Skill, work through a Workflow, and run recurring work through a Pulse when the live configuration supports it.
Where Agents live
Agents operate within Giga’s Group permission model.
The Group helps determine who can access the Agent and which shared resources are in scope.
That means an Agent’s availability can depend on:
- the Group it belongs to
- the user’s Group membership
- the Connections available in that scope
- the Skills available to the Agent
- the current live permissions
Agent instructions
An Agent can have durable instructions that shape how it behaves across repeated work.
Those instructions can define things such as:
- its role
- its responsibilities
- preferred ways of working
- constraints
- terminology
- expected output style
Keep durable Agent behavior in the Agent’s instructions rather than re-explaining it in every Chat.
Learn about Agent instructions →
Agents and context
Agents can also have durable context relevant to their role.
The live Giga environment determines what context and resources the Agent can actually retrieve.
Giga should never assume an Agent has access to every Track, Connection, file, or piece of company context simply because those resources exist in the workspace.
Learn about Agents and context →
Agents and Connections
Connections let an Agent work with live external systems.
For example, an Agent might be allowed to use:
- a CRM
- an analytics platform
- a support tool
- an internal database
- an MCP server
The Agent can only use Connections that are available within its current configuration and permission scope.
Learn about giving Agents access to Connections →
Agents and Skills
A Skill gives Giga reusable instructions for a specific kind of task.
An Agent can use Skills when the live Agent and Skill configuration makes those Skills available.
For example:
Sales Agent
Maintains the broader role, context, and access needed for sales research.
Account research Skill
Defines a reusable method for researching an account.
This keeps the Agent’s identity separate from the reusable task method.
Agents and Pulses
The current Pulse model can target an Agent.
That means recurring scheduled work can run through a named Agent when the live Pulse configuration supports it.
For example, a reporting Agent could run a recurring weekly check through a Pulse.
Learn about Pulses with Agents →
A simple example
Imagine your Sales team creates a Sales Research Agent.
Role
Research target accounts and prepare useful sales context.
Context
Uses the durable background relevant to the sales team.
Connections
Can use approved sales tools when permissions allow it.
Skills
Uses repeatable research methods without rebuilding them each time.
The team gets a consistent AI role they can return to, while access remains controlled by the workspace and Group permission model.
Image placeholder
Diagram showing an Agent at the center with Role + Instructions, Context, Skills, Connections, Group permissions, and Pulses around it.
Explore Agents
Creating an Agent
Set up a reusable Agent for a role your team returns to.
Agent instructions
Define how the Agent should behave and work.
Connections
Give an Agent access to approved external systems.
Agents and context
Understand what durable context an Agent can use.
Sharing Agents
Control who can access an Agent through Giga’s permission model.