Product5.0 AI AgentsAlpha

Three ways to put AI to work.

BIA converses. Internal agents are in alpha. Codex, Claude Code, and your own agents can connect through the currently available MCP tools.

Start for free Talk to us

Shared context Visible status Scoped MCP

Three ways to put AI to work. — Product captureDemonstration dataBYOA — Bring Your Own Agent: connect your own agent.

Designed as a system

Each capability starts from the same work and retains recognizable references to its context. AI, permission, and activity coverage is stated by status.

07 core capabilities

Agents inside the work

Not another invisible automation. Visible responsibilities.

BIKLABS gives each configured agent a recognizable place: an assignment, bounded context, and activity the team can review according to the available capability.

01Identity

Know who is acting.

Supported surfaces show a name, role, and queue to distinguish the agent from an anonymous process.

02Scope

Context is granted, not assumed.

A supported run uses the projects, tools, and knowledge bounded by its configuration.

03Evidence

Activity where available.

The run surface can show outcome, activity, and metrics when the run and provider expose them.

Autonomy with boundaries

The agent acts. The system stays in control.

The operating model assigns work with identity, permitted context, and review. Coverage expands gradually while internal agents, MCP, costs, and gates continue through validation.

BIKLABS / OPERATING MODEL04 STEPS

Receive

A work item defines the goal and scope.

Read

It accesses only permitted context.

Execute

It uses only the enabled tools.

Deliver

The result and available activity arrive for review.

Visible accountabilityActivity prepared for review

Supported events help review actor, state, and outcome without presenting an exhaustive audit trail.

Three ways to work

Choose how AI joins. The context stays.

BIA is in Preview, internal agents are in Alpha, and BYOA (Bring Your Own Agent) uses the current MCP surface. They share the project model, but each door keeps its own scope.

5.1

BIA — Ask your workspace anything

Preview

BIA consults permitted project, work-item, and page context to answer and propose supported actions. Coverage depends on the enabled scope and does not assume workspace-wide access.

BIA — Ask your workspace anything — Product captureDemonstration data

Control and visibility

Autonomy that leaves evidence.

Supported surfaces bring identity, approval, cost, and activity beside the agent and the work. Each data point appears according to availability.

5.4

Agent identity — A recognizable actor

Alpha

Agent surfaces expose a name, role, and configuration so automation is not anonymous. Full integration with assignment, permissions, and activity continues to evolve during the alpha.

BIKLABS / CAPABILITYILLUSTRATIVE MODEL5.4

Inside the system

A visible name and role in the same assignment lists as the team

Permissions and context bounded for each agent identity

Available activity associated with the identity

Model structureIllustrative

5.5

Approval gates — Review where it matters

Pilot

Gates can hold supported actions before execution and request a human decision. Coverage by agent, project, and action type is under validation; we do not claim every output is blocked automatically.

Approval gates — Review where it matters — Product captureDemonstration data

5.6

Cost tracking — Tokens, runs, budgets

Preview

The runs view can show duration, tokens, and cost when the provider returns those metrics. Limits and aggregates remain under validation, and not every run exposes the same level of detail yet.

Cost tracking — Tokens, runs, budgets — Product captureDemonstration dataBYOA — Bring Your Own Agent: connect your own agent.

5.7

Run activity

Preview

Supported run events leave visible activity for reviewing actor, state, and outcome. Exhaustive provenance, export, and complete audit coverage remain in development.

Run activity — Product captureDemonstration dataBYOA — Bring Your Own Agent: connect your own agent.

Keep exploring

One system, not another silo.

Work Items

Assign work items to your agents.

Gates

Configure which actions require human approval.

BIA

BIA as a built-in workspace copilot.

Wiki

Knowledge connected to human and agent work.

Frequently asked questions

The important details, plainly.

Direct answers about how this capability fits into the work system.

01What is the difference between BIA, an internal agent, and an external agent?

BIA is the workspace conversation interface in Preview. Internal agents are in Alpha for bounded tasks. External agents such as Codex, Claude Code, or your own can use the current MCP surface while keeping their execution environment.

02Can an agent access the entire workspace?

Full access is not assumed. The agent uses the scope and tools that are enabled; the permission model and its coverage continue to expand.

03How is a sensitive action controlled?

In supported workflows, a pilot gate can pause a transition for review. The responsible person sees the proposal, actor, and context before the workflow continues.

04Can I see the cost of each run?

The Preview view shows consumption, duration, and cost when the provider returns those metrics. Coverage, aggregates, and budget limits are still under validation.

From work to review, with context

Stop managing AI.Start managing with AI.

Projects, knowledge, and AI capabilities sharing the same operating context.

Start for free See pricing