Free
$0 / month
Meet your crew. No card, no end date.
- 2 agents
- 1 working at a time
- 500 turns a month
- 1 schedule, every 6 hours
- 100 web searches a month
- 100 MB workspace, chats kept 30 days
THE QUARTERMASTER'S LEDGER
Simple pricing. You bring the keys.
We run everything else. Three plans, priced per account, and not one cent of markup on your tokens.
You pay us for the platform: the always-on engine that keeps missions alive, the crew system, the plugin sandbox, the schedules, the chat connections we hold open, the 3D companion. Your AI model provider (OpenAI, Anthropic and friends) bills you directly, at their price. We never touch that bill.
Plans
Solo
$24 / month
Your own crew, working for real.
- 5 agents
- 3 working at once
- No monthly turn cap
- 10 schedules, down to every 15 minutes
- 1,000 web searches a month
- 1 GB workspace, chats kept 90 days
- Up to 2 sandboxes, from $1.25 a day
- 7 days free to start
Entrepreneur
$69 / month
A full crew, all working at once.
- 25 agents
- 10 working at once
- No monthly turn cap
- The longest conversations your model can handle
- 50 schedules, down to every 5 minutes
- 3,000 web searches a month
- 10 GB workspace, chats kept 180 days
- Up to 3 sandboxes, from $1.25 a day
- 7 days free to start
No card for Free. No end date. Solo and Entrepreneur start with 7 days free (card required, first paid plan only). Cancel any time and drop to Free: nothing is deleted.
Every plan gets the lot: bring your own model keys · every plugin · crews and delegation · companion mode · Telegram, WhatsApp, Slack and plugin channels · your own servers over SSH or tunnel · 45 built-in tools.
Billed monthly in USD by card through Stripe.
EVERY BARREL COUNTED
Every limit, in plain numbers.
No asterisks, no fine print. This is exactly what each plan gets.
| Limit | Free | Solo | Entrepreneur |
|---|---|---|---|
| Price per month | $0 | $24 | $69 |
| Agents | 2 | 5 | 25 |
| Chats working at once (specialists and schedules count) | 1 | 3 | 10 |
| Steps per run (a model call and its tool calls) | 8 | 12 | 20 |
| Turns per month (one turn = one run of an agent) | 500 | No cap | No cap |
| Context per run | 100k tokens | 400k tokens | The model's full window |
| Web searches per month (built-in search) | 100 | 1,000 | 3,000 |
| Schedules (regular check-ins included) | 1 | 10 | 50 |
| Shortest schedule interval | Every 6 hours | Every 15 minutes | Every 5 minutes |
| Live connections (chat sockets held for plugins, e.g. Discord) | 1 | 5 | 20 |
| MCP servers | 1 | 5 | 20 |
| Plugin wakes per hour | 60 | 600 | 3,000 |
| Workspace storage (whole account) | 100 MB | 1 GB | 10 GB |
| Workspace storage (per agent) | 64 MB | 256 MB | 1 GB |
| Sandboxes (private computers for your agents, paid from your wallet) | — | 2 | 3 |
| Idle chats kept for | 30 days | 90 days | 180 days |
| Files per message | Image 5 MB · audio 16 MB · document 10 MB · up to 10 | Same | Same |
Over your working-at-once limit, work simply waits its turn. Nothing fails.
The step limit never stops the work. When a run uses its steps, the agent goes back in the queue and carries on from where it was, the way an operating system shares time so no single app takes the whole machine. A long mission runs as long as it needs.
NO CUT OF YOUR TOKENS
Your model, your price, your bill.
We make money on the platform, never on your AI usage.
Your provider, your rate
Connect OpenAI, Anthropic, OpenRouter and many more, a free model on your own computer through the tunnel, or a provider plugin. They bill you. We never see that invoice.
See where the spend goes
The Usage page estimates cost per chat, per agent, per model and per schedule, from the prices you enter. Export it as CSV.
Plan changes, priced up front
Changing plans shows you the exact prorated amount before you confirm. A downgrade that doesn't fit your current usage tells you which limit is in the way.
TRIMMING THE SAILS
Cheap models where they fit. Zero spend while waiting.
Every part of the platform is built to spend fewer tokens.
Waiting is free.
An agent waiting on a build, a timer or a colleague makes no model calls until something arrives.
Prompts built for the cache.
The part of every prompt that repeats is sent exactly the same way each time, so your provider bills it at its cheaper cached rate.
A model per specialist.
Put a fast, cheap model on research and your best one on the lead. You pay flagship prices only where they count.
No double charge after a restart.
A saved reply is delivered without a second model call.
No silent swaps.
The model you pick is the model that runs; we never switch to another behind your back.
Watch it live.
Each specialist's card shows its running cost, with a Stop button.
ASK THE QUARTERMASTER
Straight answers, no fine print.
The questions people ask before they sign up.
Can I use my ChatGPT Plus or Claude Pro subscription?
Yes, through our tunnel. Log in to OpenAI's Codex with your ChatGPT plan on your own computer, connect that computer with the tunnel, and your agents can think with your subscription. Claude Pro and Max work the same way through Claude Code, using your plan's extra usage. Subscription models are text only. Prefer to pay per use instead? Paste an API key from the provider's developer platform: Get an API key in five steps.
What will my model bill be?
It depends on the models you pick and how much your crew works. Put a fast, cheap model on research and your best one on the lead. An agent that's waiting makes no model calls, and the Usage page estimates spend per agent and per chat. Want a hard ceiling? Set a monthly limit on your key in your provider's dashboard. Get an API key in five steps.
Why does AgentParley use fewer tokens?
Every time an agent thinks, it re-sends its instructions, its tools and the conversation so far. Model providers charge far less for input they've just seen, called prompt caching, but only if the start of the prompt is exactly the same as last time. AgentParley keeps that start identical: no clock at the top, history that only grows at the end, and a long chat summarised once instead of trimmed a little every turn. A waiting agent makes no model calls at all, and each specialist starts with a short, focused brief, not the whole conversation. Example: an agent re-sends a 30,000-token start 20 times on a task. Uncached, that's 600,000 full-price input tokens. At typical cached-input pricing (about a tenth of the normal rate), 19 of those 20 are read from cache, roughly 85% less on the repeated part.
Do I need to know how to code?
No. You describe each agent in plain words, who it is and how it works, and pick its tools from a list. The console assistant can draft an agent for you.
Do I need a server or anything installed?
No. AgentParley runs in our cloud. No hardware to buy, no server, no Docker, no terminal. Sign up in your browser and your first agent is already there. If you do have a machine you want agents to work on, or a local model you want them to use, connect it in a few clicks.
How long does setup take?
Create an account, paste an API key from the AI provider you pick, and send your first message. The console walks you through it, and you can skip ahead any time.
Where do I get a model key?
From the developer platform of the AI provider you choose: OpenAI, Anthropic, OpenRouter and many more. Every plan is bring-your-own-key, so you pay that provider directly at their price. Web search is included up to your plan's monthly allowance.
Are my keys safe?
Keys are encrypted with a key unique to your account, kept in Amazon's key vault. After you save a key, the console only shows 'set' or 'not set'. Nobody can read it back out, including you. More on the Security page.
Do you charge for tokens?
No. Your provider bills you. We take no percentage and never sit in that payment.
What's a crew?
A group of AI agents that work together. One lead agent takes your request and hands pieces to specialists: say, a researcher on a fast cheap model and a writer on a top model. Each specialist reports back, and the lead only answers you once every report is in. You decide which agent may talk to which. See Crews.
Do subagents cost extra?
Not on our side. They use tokens from your provider like any chat, count toward your working-at-once limit, and show up separately on your Usage page.
What is a sandbox?
A private computer in the cloud that only your agents use, where they build, run and keep things apart from your own machines. Small is $25 a month or $1.25 a day, Large is $45 a month or $2.50 a day, paid from your wallet. Solo has room for up to 2 and Entrepreneur for up to 3. See Sandboxes.
What's companion mode?
Give any agent a 3D face and a voice. You talk to her in the web console, she answers out loud with lip-sync and gestures, and she can hand real work to your crew. Voice uses your own key from ElevenLabs, fish.audio, Deepgram, an OpenAI-compatible vendor, or a voice plugin. Without one, she still works, with captions. See Companion.
Are plugins safe to install?
Every plugin call runs in a fresh, throwaway sandbox with no access to files or processes. It can make web requests to the addresses it declared, and ask your agent's model to think through our API without ever seeing your API key. Before you install, you see what it hooks into, its permissions and every website it may call. If an update asks for more access, it waits until you approve it. More on the Security page.
Do plugins cost extra?
No. Every plugin is free to install on every plan. Plugins that listen to chat platforms count toward your plan's plugin wakes per hour.
Can I cancel any time?
Yes. Cancel from Plan & billing and you drop to Free at the end of the period you paid for. Moving plans deletes nothing. What you already have stays; you just can't add more until you're back under Free's limits. Free's 30-day chat retention applies from then on.
How do I delete my data?
Delete your account yourself from Account settings. It locks straight away, and you have 30 days to change your mind and restore it. After that, everything is purged. You can export any chat as a transcript, download your workspace files, and export usage as CSV before you go.
Is there a trial or annual billing?
Yes to the trial: your first paid plan, Solo or Entrepreneur, starts with 7 days free. It needs a card, and you can cancel before the week is up and pay nothing. The Free plan also stays free with no card and no end date. Paid plans are billed monthly, by card, through Stripe.
Can I self-host AgentParley?
No, and that's the point: we run it so you don't have to. If self-hosting is a must, OpenClaw and Hermes Agent are good open-source options, and we say so on our comparison pages: OpenClaw alternative, Hermes Agent alternative.
Is AgentParley a hosted OpenClaw?
No. AgentParley is its own hosted platform, not OpenClaw run for you. It does the same kind of job without a machine to run, and adds standing crews, a 3D companion and sandboxed plugins. See the OpenClaw comparison.
For power users
Which models can I use?
Anthropic directly, plus anything that speaks the OpenAI API: OpenAI, Azure, Groq, Together, OpenRouter, DeepSeek. Pick a different model per agent. A plugin can add a new provider.
Can I use local models?
Yes. Ollama, LM Studio, vLLM or llama.cpp on your own box, through our tunnel, with no port to open.
Do you support MCP?
Yes. Connect remote MCP servers from the console and their tools show up for your agents next to the built-in ones. Free allows 1 server, Solo 5 and Entrepreneur 20. Plugins are the other way to add tools, model providers and chat apps, each call in its own sandbox.
Can I cap what I spend?
Set a monthly limit on your key in your provider's dashboard. On our side, every chat shows its running cost with a Stop button, and the Usage page breaks spend down per agent.
What's a step?
One model call plus the tools it asks for. The step limit doesn't block your agent: when a run uses its steps, the work goes back in the queue and picks up where it left off. It's how an operating system shares time between apps, so one busy agent can't take everything. Free allows 8 steps per run, Solo 12 and Entrepreneur 20.
Can my agents use the web and my computers?
They search and read the web, and on servers you register they run commands over SSH or the tunnel.
Does my agent remember?
It keeps the chat history, and summarises it automatically when it gets long, plus its own workspace files and a shared crew folder. A plugin can replace the summariser.
Can I take my stuff with me?
Export any chat as a transcript, download your workspace files and export usage as CSV.
Will updates break my setup?
Platform updates ship on our side. Plugins you've pinned stay on their version, and a plugin update that asks for more access waits for your approval.
Is it open source?
The platform is hosted, not open source. Plugins are TypeScript, and our official plugins are readable examples on the Developers page.
Start free. Upgrade when it pays.
Start on Free, no card, and move up when your crew outgrows it.