Agent.Space has two parts: dependable model capacity and a persistent place where that capacity becomes finished work.
01UP TO 50% OFF
Model plans
A simpler way to subscribe to models
One credits balance for every model: pay-as-you-go channels need no purchase and charge credits as you call, and plans save more on the models you use most. Plans, credits, billing and Model API Keys stay in one place.
Codex, Claude Code, and other Agents keep running in the cloud. Sessions, files, previews, and team context remain together so work can continue at any time.
One credits balance for every model: pay-as-you-go channels need no purchase and charge credits as you call, and plans save more on the models you use most. 1 credit = US$1.
1
Add credits
Top-up credits never expire, or start a credit subscription that tops up every month. 1 credit = US$1.
2
Pay as you go
Every pay-as-you-go channel is ready to use with no purchase. Calls are charged to credits by the tokens actually used.
3
Save with a plan
Subscribe to a plan for the models you use most. Its allowance goes first; once it runs out, or for models outside the plan, calls switch to pay-as-you-go automatically.
New model
GPT-6 Astra is now available
GPT-6 Astra is included in every Codex Dedicated plan
Codex DedicatedFixed weekly allowance · paid from credits monthly or yearly
Coming soon
More official subscriptions, shared
SuperGrok and Claude Code group plans are in preparation. Preview the benefits now.
Coming soon
xAI · Grok
SuperGrok
Grok 4 with far higher limits, Imagine and early access
Opening soon
Grok 4
Now available
Plan benefits
Grok 4 flagship with much higher usage limits
Grok Imagine image and video generation
Real-time X and web search
Early access to new models and features
Independent account, no shared login
Your own API key and model API access
Coming soon
Anthropic · Claude Pro
Claude Code
The terminal coding agent, with Pro usage shared through the group
Opening soon
Claude Opus 5
Claude Code ready
Plan benefits
Claude Code takes over your repo end to end
Claude Pro usage with Opus models
Use your local Claude Code client
MCP plugins and custom Skills
Keep project files and conversations
Independent account, no shared login
Plans are paid from credits each period. Annual plans take the full year at once while the weekly allowance still resets every 7 days. Top-up credits never expire; credit subscription credits last for their period.
Resources matched automatically
Clear capacity levels based on official Codex allowances.
Allowance stays isolated
The platform keeps allowance and access boundaries separate.
7-day cycles, clear rules
Allowance resets in full every 7 days. Unused allowance does not roll over.
Independent account and Workspace
Plans, Agents, files, and collaboration stay in one place.
Pay-as-you-go channels
Official pay-as-you-goPay from credits for what you use
One credits balance runs GPT, Kimi, GLM, MiniMax, DeepSeek, Claude, and Grok pay-as-you-go in Agent.Space or through the Developer API. Credits are charged only for measured model usage, and top-up credits never expire.
Every plan and pay-as-you-go channel, side by side
No sign-in needed: compare each model across channels and plans by price, privacy and capability. Pick a pay-as-you-go channel and start using it right away.
Use Agent.Space plans and credits in any Agent product
Codex Dedicated plans cover OpenAI models, and credits pay for other available models as you go. Use them in Agent.Space, Codex, Claude Code, OpenCode, Hermes, and other Agent products you already work with.
Use it directly in Agent.Space. For other products, connect your Developer API once and keep working there.
GPT Image 2 uses your plan allowance first, then credits when no plan is available or its allowance is used up. Other models use their listed funding source.
Use Agent.Space models in the client you already love
Download Codex, Claude Code, or OpenCode and point it at the Agent.Space Developer API. One API key brings your plan allowance, credits, and every available model into the terminal and editor you already work in.
Codex
OpenAI · Coding agent for terminal and IDE
OpenAI's official coding agent with a CLI, IDE extensions, and a desktop app. It connects over the OpenAI Responses protocol and runs the OpenAI models in your Codex Dedicated plan.
macOS
Linux
Windows
Install · macOS / Linux
bash
curl -fsSL https://chatgpt.com/codex/install.sh | sh
Anthropic's official coding agent, speaking native Anthropic Messages. Point its base URL at Agent.Space to run Claude models pay-as-you-go with credits.
An open-source terminal agent with custom providers, compatible with both OpenAI and Anthropic protocols. One config unlocks every model available on Agent.Space.
OpenAI, Anthropic, DeepSeek, and more share a single key and endpoint. Switch models without switching accounts.
Two protocols
OpenAI Responses and Anthropic Messages
Connect over each client's native protocol. No proxies and no client patches: configure once and keep working.
Clear billing
Plans and credits stay separate
Models in your plan use its allowance first, the rest is paid from credits, and every request is visible in Request history.
Your choice
Swap clients any time
Codex today, Claude Code tomorrow. Your key and allowance stay the same, so your workflow never breaks.
The template already includes the endpoint and selected model. Replace <your-api-key>, then keep the config file local and never commit or share it.
04 / Cloud Workspace
One cloud Workspace for every leading Agent
Codex, Claude Code, and other leading Agents share one cloud Workspace. Files, context, and progress persist so individuals can start immediately and teams can take over at any time.
New chat
Which Agent should we start with?
Multiple Agents and teammates share one Workspace to move real work forward together.
Import local conversations
Codex — describe what you want to build, fix, or review…
Each Session keeps its Agent, files, and context in the same Workspace, so people and Agents can hand work off without starting over.
People and Agents, moving work forward together
01
Multiple Agents move the work forward
Codex, Claude Code, OpenCode, and other Agents can take on separate tasks while sharing the same files and project context.
02
Everyone sees the same working context
Invited teammates can see the same Sessions, files, and progress without repeatedly explaining what has already happened.
03
Change people or Agents without restarting
Context, code, results, and Previews stay in the Workspace, ready for the next teammate or Agent to continue.
04
Work, access, and progress stay clear
Roles, permissions, task status, and usage are visible in one place, so everyone knows who owns the next step.
05 / Security and trust
Your work stays in your Workspace
Your Workspace is separate from everyone else's. Only people you invite can see its files and tasks.
We use this content only while Agents do the work you ask for—not for advertising.
Independent runtime
Isolated Workspace
Your Workspace
Files · Sessions · Permissions
No shared accounts
Managed access
Team permissions
Invited members only
Task data
Used only for the work
↓
Files and results
Export anytime
01
Independent runtime
Every Workspace runs independently
Projects, files, and Sessions run in an isolated Workspace environment, separate from other work.
02
No shared accounts
No account or password exchange
The platform manages service access and resource matching; users never share upstream accounts or login credentials.
03
Team permissions
Only invited members can join
Invite people into the Workspace and control their access with clear roles and permissions.
04
Export anytime
Take your files and results with you
Code, files, and results stay in ordinary directories and standard formats, ready to export when needed.
You control who can access the Workspace, how work runs, and when the project leaves.
06 / FAQ
Straight answers before you start
The essentials about Codex Dedicated, plan allowance, credits, and collaboration.
Agent.Space Blog
Notes for building with AI Agents
Practical guides, product updates, and field notes for putting cloud Agents to work.