AgentWorks
Try the demo Book a call

Solutions

GTM Ops Customer support Finance & billing Shopify Ops Marketing & growth Growth & Customer Analytics Engineering Ops All premade agents

Platform

Overview Goals Crews Studio Flows Brain Vault LLM Gateway AgentBridge How we work Blog Try the demo Book a call

Product

One platform. Agents inside your process.

For your teams: Goals work toward a number you set and measure every run, Crews turn your team's know-how into experts anyone can ask, and Flows run fixed steps the same way every time. For IT and leadership: Brain is the shared knowledge every agent reads, Vault governs every MCP tool and shared secret, LLM Gateway (coming soon) controls who can call which model, and Studio gives every employee a ready AI workspace with cost and usage visible to the company. AgentBridge connects everyone's personal AI, like Claude, ChatGPT, Cowork and Cursor, to the company's agents. All of it runs on the AI plan you already pay for.

Book a call Watch the 60-second demo

For your teamsGet the work done.

GoalsGive an AI agent a goal and a metric. It works toward the target and measures every run.CrewsBuild an expert agent once. Your team asks it from Slack, Claude, ChatGPT or Cursor.FlowsA fixed chain of agents you run from anywhere.

For IT and leadershipStay in control.

BrainShared company knowledge: write it once and every agent reads it.VaultOne governed door to every MCP tool and shared secret.LLM GatewayComing soon: manage model access and see the cost per person.StudioThe online workspace where you build agents and Crews, with skills, tools and integrations ready on day one.
60-second demo: an agent with the goal "Book more sales demos". Illustrative data.

Scripted where it counts

Not everything should be agentic. Script what's known. Use agents for judgment.

Pulling data, calling an API or applying a change should happen the same way every time. AgentWorks runs that work as scripted steps: plain code with no AI in the loop, checked before the workflow moves on. Agents take the steps that need judgment, and the two hand off through files and data, not chat.

Refund reviewIllustrative

Runs every morning

  1. ScriptPull yesterday's refund requestsAPI call, no AI
  2. AgentCheck each against the orderjudgment
  3. RuleOver $200?decided in code
  4. ApprovalYou approve the big oneshuman step
  5. ScriptIssue the refundssame way every time
Script: code, no AIAgentRule in codeHuman approval

  • Same result every time

    A scripted step runs as code and must exit cleanly and pass its output check before the next step starts. If it fails, the run stops and says why, instead of guessing.

  • Cheaper and faster

    No model tokens for work that doesn't need thinking. Agents spend your AI plan only where judgment moves the result.

  • Written for you, readable by you

    Describe the step and the Builder writes the script and tests it before it runs. The code is yours to read, review and change, and every run is logged.

Under the hood

Three layers, built to run for months.

  1. 3Workflows & goalsPipelines of deterministic and agentic steps, learnings and a knowledge base, measured against a goal
  2. 2CrewsAn agent with skills, memory, a browser, Slack and WhatsApp, triggers, and calls to other crews
  3. 1AgentsVendor-native Claude Code, Codex, Cursor, Pi and Muse in live terminals, with your MCP tools and sandboxed commands

Layer 1 · Agents

Real agents, not a model behind a wrapper.

Most AI tools call a model and bolt their own loop around it. AgentWorks runs each vendor's own agent, with the model and the harness its makers built and tuned together, so you get everything the agent can do.

Your goalDemos booked per week · target 5AgentWorks plans, schedules, measures and improves
Sandbox for every command · only the folders this workflow was granted
Native agent in a live terminal
  • Claude Code
  • Codex
  • Cursor
  • Pi
  • Muse
Model + harness from the company that built it · watch it live, step in anytime
MCP serversYour tools and AgentWorks' own
Shell & CLIsgit, gh, cloud CLIs, scripts
BrowserIts own persistent, isolated session

  • The full agent, not just the model

    Claude Code, Codex, Cursor, Pi and Muse run as themselves in a live terminal session, with the planning and reasoning their makers built. That is also why they run on the subscription you already pay for.

  • Your tools, governed

    Every agent gets your MCP servers, AgentWorks' own tools, a shell and a browser, granted per workflow. Writes go through AgentWorks' tools, so every change is checked and logged.

  • Sandboxed where it acts

    Every command and script an agent runs executes in an OS-enforced sandbox (Landlock on Linux, sandbox-exec on macOS), limited to its workflow's folders with separate read and write access, and with its own private /tmp and home folder.

Bring your own AI

Use the AI plan you already pay for.

AgentWorks runs on your Claude, ChatGPT, Gemini or Cursor subscription. No credits, no per-task fees, no marked-up tokens.

  1. Connect your planSign in once with Claude Code, Codex, Cursor, Pi or Muse, using your subscription or that CLI's API key.
  2. Match the model to the jobYour strongest model for judgment, a cheaper one for routine runs.
  3. Keep working at the limitWhen a plan hits its limit, work pauses and resumes at the reset, or moves to another plan you've connected.

Connected AI plans3 plans · 4 teammates

Claude MaxJudgment, planning, customer-facing draftsIn use
ChatGPT ProResearch, reports, routine runsIn use
GeminiTakes over when a plan hits its limitStandby

Under the hood

Built on the tools you already trust.

  • Your AI plan

    Runs each vendor's own coding agent (Claude Code, Codex, Cursor, Pi or Muse) on the subscription you already have, or on that agent's API key.

  • Its own browser

    A persistent, isolated browser per workflow, so agents can work in any web app you use.

  • MCP and APIs

    Connect any MCP server or API. Tools are granted per workflow, not globally.

  • Secrets vault

    Encrypted, injected only at run time, never shown in chat or logs.

  • Sandboxed commands

    Every command an agent runs gets an OS-enforced sandbox (Landlock on Linux, sandbox-exec on macOS) and a private /tmp and home folder.

  • Full run log

    Every step, tool call, decision and cost, kept per run.

FAQ

How it works, answered.

Does every step use AI?

No. Work that should run the same way every time, like pulling data, calling an API or applying a change, runs as a scripted step: plain code with no AI, whose output is checked before the workflow moves on. Agents take the steps that need judgment, and rules like "over $200 needs approval" are decided in code.

What's the difference between a Goal and a Crew teammate?

A Goal owns an outcome and keeps working on its own schedule. A Crew teammate is who you talk to: you hand it a job in Slack or WhatsApp, or ask how a goal is doing.

Can I use it from ChatGPT or Claude?

Yes. AgentWorks includes an MCP server. Connect it to ChatGPT, Claude, Cowork or any MCP client to check goals, read reports and start runs from that chat. Hosted AI apps connect to your AgentWorks server with a sign-in link.

Does it work with the tools I already use?

Yes. Agents use MCP servers, APIs and their own browser, so anything you can do in a web app, they can do too, with credentials kept in the vault.

Start with one goal or one expert. We'll set it up with you.

We set up your first agents with your team, agree how success is measured and connect the tools they need.

Book a call Try the demo