Define the job once, in plain instructions.
Set the agent's role, tone, and rules as instructions so it handles the task the way your team expects, every time.
DelegateInstruction-based agent definition
PRODUCT · AGENTS
Package repeat work into an agent your team trusts.
An agent combines your instructions, the knowledge it is allowed to use, and the actions it may take into one AI teammate for a specific task. Share it across the team so everyone gets the same cited, permission-aware behavior without re-explaining the job.
Starts withA task your team repeats
Grounds inInstructions plus scoped connected knowledge
ProducesA shared agent that answers and acts in scope
BoundaryAn agent only uses the knowledge and actions you grant
THE PRIMARY JOB
Repeat questions should not need repeat explanations.
Agents package a task's instructions, the exact knowledge it should rely on, and the actions it is allowed to take into one reliable teammate. An onboarding helper, an engineer copilot, or a legal reviewer behaves the same way every time, grounded in the sources you chose.
- 01
The same question is answered a different way by each person.
- 02
A reliable answer depends on knowing which sources to trust.
- 03
Repeat work is re-explained instead of reused.
SEE IT WORK
The Onboarding agent answers from the current HR and IT docs with citations, then prepares the access request for review.
LAgentsProject AtlasShared · Engineering
Onboarding Assistant · custom agent
Answers first-week setup questions from approved HR and IT docs, and prepares access requests for review.
- Knowledge
- 3 scoped sources
- Actions
- 2 tools, 1 action
- Shared
- Engineering
Confluence
SharePoint
DriveInternal searchBuilt-in toolWeb searchBuilt-in toolAccess requestCustom action · OpenAPI
Sample ask · How do I get repo access?
The agent answers from the IT runbook with citations, then prepares the access request and holds it for IT review.
HOW THE PRODUCT WORKS
Three visible parts of Agents.
Scope the knowledge it can use.
Attach only the connectors, document sets, or files the task needs. A narrower scope produces more reliable answers.
DelegateScoped connector and file knowledge
Grant only the actions it needs.
Choose which built-in tools and custom actions the agent may use, from internal search to governed writes in connected systems.
DelegatePer-agent action allow-list
Share it with the people who need it.
Publish an agent to individuals or user groups so the whole team runs the same vetted setup.
DelegateTeam and user-group sharing
AGENTS TO ACTIONS
An agent acts through governed actions.
When a task needs a change in a connected system, the agent uses only the actions you allowed, prepared and held for review where required.
Explore ActionsVerified context
A shared agent that answers and acts in scope
ActionsA task that needs a change in a system
AGENTS CONTROL BOUNDARY
The product handles the work. Your team keeps the decision.
Linkence handles
- Hold the agent's instructions and scope
- Answer only from the granted knowledge
- Use only the allowed tools and actions
- Apply the same behavior for every team member
Your team controls
- The agent's instructions and tone
- The connectors, sets, and files in scope
- The tools and actions the agent may use
- Who the agent is shared with
Not designed for
- Unscoped access to every connected source
- Actions the agent was not granted
- Replacing human accountability for a decision
AVAILABILITY AND LIMITS
availableWhat is available, and where the boundary sits.
Required setup
- An administrator or member with permission to create agents
- The connectors or files the agent should rely on
- The allowed tools and any custom actions configured
Known limitations
- An agent is only as reliable as the knowledge scoped to it.
- Allowed actions depend on admin-configured tools and connector capability.
- Sharing and access follow the workspace's user-group and admin controls.
An agent's knowledge can be scoped to chosen connectors, document sets, or files.Verified product claimAn agent's available tools and actions can be restricted to an allowed set.Verified product claim
CONCRETE WORKFLOWS
Three ways teams use Agents.
Onboarding assistant
A new joiner's setup question
ConfluenceSharePointDrive
A cited answer plus a prepared access request
IT reviews and approves the requestEngineer copilot
A question about a service or incident
GitHubJiraConfluence
A cited answer scoped to the team's repos and runbooks
An engineer decides the formal changeLegal reviewer
A clause or policy check on a document
SharePointDriveConfluence
A cited comparison against current policy
Counsel makes the legal judgmentCONTINUE THE WORK
Products and systems connected to Agents.
AGENTS QUESTIONS
Direct answers about Agents.
Availability and connector-specific behavior are stated directly, without universal claims.
Can an agent see everything our company connected?+
No. An agent only uses the connectors, document sets, or files you attach to it, and it still respects each person's source permissions when it answers for them.
What can an agent actually do?+
Only what you allow. You choose the built-in tools and custom actions it may use, from internal search and web search to governed writes in connected systems.
Can we share one agent with the whole team?+
Yes. Agents can be published to individuals or user groups so everyone runs the same instructions, knowledge scope, and allowed actions.
SEE AGENTS ON YOUR DATA
Start with one real question or workflow.
Connect the relevant sources and show Linkence the approval rules. See the path from evidence to completed work on your own data.
Representative data is enough for the first demo. No production access required.