The web console
Chat, watch every specialist work live, and steer mid-mission.
CHARTING THE COURSE
From sign-up to a working AI crew.
Four steps, all in your browser. Nothing to install, nothing to host, nothing to update.
Self-hosted agents mean a machine that stays on, a terminal, an install script and updates you run yourself. AgentParley is the hosted version. We run the servers, the schedules, the retries and the sandboxes. You bring an API key and an idea of what you want done. No code: you describe each agent in plain words.
THE ROUTE
Four steps. No terminal.
- 1
Sign up in your browser
Email and password, or sign in with Google or GitHub. The Free plan needs no card. Your first agent is already waiting when you land.
- 2
Bring your own model key
Paste an API key from the AI provider you pick: OpenAI, Anthropic, OpenRouter and many more. Or use a local model on your own machine through our tunnel (see below). Your provider bills you directly, at their price. We add no markup. Then send your first agent a message and watch it answer.
New to API keys? Get one in five minutes from OpenAI's API platform or Anthropic's Console. Get an API key
- 3
Build your crew
No code. Give each agent a soul (who it is), instructions (how it works), a model and the tools it may use. Put specialists under a lead and decide who may call whom. Not sure where to start? Ask the built-in assistant in the console to draft an agent for you, and approve it before anything changes.
- 4
Point them at the work
Ask in a chat, set a schedule, or hand an agent a webhook your tools can call. It writes itself a to-do list and keeps going until every item is done, even after you close the tab.
HAILING FREQUENCIES
Talk to your crew where you already are.
The same agent, whichever way you reach it.
Your chat apps
Telegram, WhatsApp and Slack are built in. Discord is one plugin away, and plugins can add more. You decide who's allowed to talk to each agent.
A companion
Give an agent a 3D face and a voice in the web console. Talk out loud; she hands the work to the crew and tells you the result. Meet the companion
NO SHIP TO MAINTAIN
No server, no install, no updates.
The chores are ours: no server bill, no update night, no machine humming in a closet.
Nothing to install
No download, no Docker, no Node or Python. A browser is enough.
No server
Your crew runs on our servers in the US, awake around the clock, whether your laptop is open or not, so you never keep a computer on for it.
Automatic updates
The platform improves under you. There is nothing to upgrade, ever. Plugins can update themselves too, or stay pinned to a version you trust, and any update that asks for more access waits for your OK.
Your model, your key
Use the provider you trust. The model you pick is the model that runs; we never swap it for another behind your back.
BELOW DECKS
Agents that keep running 24/7.
The plain-language version of the machinery.
Every job is written down first.
Before an agent is woken, the work is saved. If something crashes, another worker picks it up. Nothing is silently lost.
Your crew works in parallel.
Several agents can run at once, up to your plan's limit. Over the limit, work waits its turn.
Waiting costs nothing.
An agent waiting on a build, a timer or a reply makes no model calls until something arrives.
You can see the cost.
Estimated spend per chat, per agent, per model and per schedule, from the model prices you enter. Export it as CSV.
Agents reach only what you allow.
Each agent gets the tools, servers and colleagues you choose, nothing more.
- scheduleThe weekday 7:00 schedule wakes lookout
- lookoutSearches the web for this week's packaging news
- lookoutReads the top three articles
- lookoutSaves Monday's brief to the shared folder
- lookout
send_to_channelTelegram "Morning brief: three stories worth your coffee. Full notes in the shared folder." - systemlookout rests until the next run, and costs nothing while it waits.
Set the schedule once; the brief arrives on Telegram every weekday, whether your laptop is open or not.
SHORE LEAVE
Got your own servers? Bring them.
Everything above works with nothing but a browser. If you want agents working on your own machines, connect them, or install our small tunnel on a box behind your router so no port has to be opened. Free models running on your own computer, such as Ollama or LM Studio, can power your agents through the same tunnel.
Your servers
Real work on real machines
Agents run commands on the machines you connect, through the tunnel or a direct connection. You choose which agent may use which host.
Your models
Run models at home
Free models running on your own computer, such as Ollama or LM Studio, show up as models in your account. Pick one per agent. Text only.
Your subscription
Use ChatGPT or Claude
Log in to OpenAI's Codex with your ChatGPT plan, or to Claude Code with your Claude plan and its extra usage, on the tunnel box. Your agents can then think with that subscription.
Nothing inbound
Dials out, cuts off fast
The tunnel only dials out to us, so no port opens on your network. You approve each box once in the console, and revoking it cuts it off within seconds.
Send your first agent its first job.
Sign up free, paste an API key, and your crew gets to work.