Docs

Run agents like employees.

Everything you need to design, deploy, and supervise an AI workforce.

60-second start

Skip the manual. Install a starter team.

The fastest way to learn the product is to install one of the curated starter teams from the marketplace. Each one ships with system prompts, tools, and a suggested first task pre-wired. Once it's running, peel back the layers.

  1. 01
    Sign up
    No card required. $5 LLM credit on us.
  2. 02
    Install a starter team
    3-4 agents pre-wired for a specific job.
  3. 03
    Pick the suggested task
    The form is pre-filled — just click Run.
  4. 04
    Approve risky calls
    Watch agents stream their work.

All topics

What do you want to learn?

Getting started

Go from signup to first agent run in under 60 seconds.

  • Create your account
  • Pick a starter team from the marketplace
  • Assign your first task
  • Approve risky tool calls
  • Read the run log

Agents

How agents are configured, prompted, and constrained.

  • Anatomy of an agent: role, prompt, model, tools, autonomy
  • Generating an agent from a prompt
  • Editing system prompts safely
  • Per-agent cost caps (run + daily)
  • Reading the scorecard

Teams

Group agents into pods that share goals and context.

  • Building a team manually
  • Generating a team from a prompt
  • Team types and what each is good for
  • Setting a goal that propagates to every member
  • Reading the team scorecard

Workflows

Sequence agents, tool calls, and human approvals.

  • Step types: agent_run, tool_call, approval_gate
  • Template variables in prompts ({{ workflow_input }}, {{ steps.<name>.output }})
  • Per-step approval gates
  • On-failure: abort / continue / retry
  • Watching a run stream

Tools

What agents can call — built-ins, custom HTTP, secrets.

  • Built-in catalogue (web_search, http_fetch, web_browse, code_execute, GitHub, knowledge_search, remember)
  • Risk levels and approval gates
  • Custom HTTP tools (Pro+) with template substitution
  • Secrets vault: API keys without leaks
  • E2B power tier: persistent sandbox, 4 GB RAM, batteries-included Python

Marketplace

Install pre-built agents, teams, and workflows.

  • Three template types: agents, teams, workflows
  • One-click install into any company
  • Publishing your best agents (Pro+)
  • Versioning and parent chains
  • Featured templates

Memory & knowledge

Long-term context that agents pull from on demand.

  • Business memory: scoped key facts agents remember
  • Knowledge base: PDFs, markdown, URLs — chunked + searchable
  • Self-improvement loop: agents propose memory saves
  • Decision log: traceable record of choices

Metrics & cost

Track reliability, spend, and tool failure rate.

  • Reliability score (0-100): how to read it
  • Per-agent and per-team scorecards
  • Cost rollups: today, yesterday, 7d, 30d
  • Run logs: every tool call with timing and cost
  • Cost caps: per-run, per-agent-per-day, per-user-per-month

Billing

Plans, LLM credits, and bring-your-own keys.

  • Free vs Pro vs Team — what differs
  • How LLM credits work (rollover, overage)
  • Bring your own Anthropic / OpenAI keys
  • Hard limits: what happens when credit runs out
  • Managing your subscription via Stripe Customer Portal

Deep guides are on the way.

Right now the best way to learn is hands-on: sign up, install a starter team, and watch it work. Email hello@agenticbusiness.io with questions and we'll answer + turn the answer into a doc.

Docs · Agentic Business