One active session per line
Each active Standard Code session occupies a line of service on your plan. Lines of service can be added to your plan at any time if you require additional parallel sessions.
Standard Code is a coding agent in the cloud with unlimited usage. Like a cell phone plan, you pay for lines of service. Each line on your plan powers one active session at a time.
Join the waitlist. We'll send you a key when it's your turn.
Already have an access code? Start your $5 1-week trial → Standard Code is able to offer unlimited usage because of its unique lines-of-service billing model and a novel, curated agent architecture.
Each active Standard Code session occupies a line of service on your plan. Lines of service can be added to your plan at any time if you require additional parallel sessions.
For each active line of service on your plan, you can run a Standard Code session with unlimited usage. Usage is constrained by linear time, not token quota. Standard Code is made up of a pre-configured, finely-tuned team of expert agents. Each is automatically invoked as needed, and sub-agents do not count against your active lines of service.
Every sub-agent is a specialized domain-specific agent with its own curated prompts, tools, and context. This allows us to tune the agents that make up Standard Code for maximum efficacy per dollar.
Standard Code's default agent leveraging open weight models. The world's first curated coding agent — powered by a coordinating team of domain-specific agents.
Standard Code's same domain-specific agent architecture powered by OpenAI models. Uses your existing ChatGPT Pro subscription.
Requires at least one active line of service.
How does it work? Take a look behind the scenes
Sessions executing on an always-on machine, such as a VPS, can continue uninterrupted even if the client you're using to drive the session disconnects.
Standard Code ships with first-party native clients for your terminal, the web, macOS, and iOS.
Thread history, queued messages, and even unsent drafts stay synchronized.
Standard Code exposes the same REST and WebSocket APIs its own clients use. Drive sessions from CI, internal dashboards, or your own product.
Create, message, list, and resume sessions from any language over plain HTTP. No SDK required.
POST /threads · GET /threads/:id
Stream tokens, tool steps, and the end-of-turn signal as they happen.
`⇅ wss://your-instance/threads/:id/ws`
Use a bearer token to run jobs from a push or a schedule.
Authorization: Bearer sk_standard_…
Give Standard Code an outcome. It sets a goal, breaks it into steps, and checks each one off. Unlimited usage keeps it working until the goal is complete.
Standard Code dispatches subagents for research, codebase exploration, asset creation, and more. Subagent work never crowds your session's context.
Standard Code uses a dedicated subagent to compact long-running threads. Compaction happens in flight and never s your session's work.
Ask for an icon, a hero image, a sprite sheet. The asset subagent creates it, runs its own QA, and places the file into your project.
Install Agent Skills that persist across sessions and connect any MCP server.
When a model provider fails, the session moves to another one mid-run and keeps its thread intact.
Per-tool permission gates mean the agent only ever touches what you allow.
Permission prompts follow you to whichever client you are watching, so a waiting session never blocks the machine running it.
Additional lines available, $49/month each · Secure checkout by Stripe
Hosted seats billed centrally: headcount × $49, unlimited usage.
The entire Standard Code stack on your infrastructure with your inference providers and visibility into usage through the Standard Code admin panel.
Join the waitlist. We'll send you a key when it's your turn.
Already have a key? [Start your $5 week →](/start)
Questions while you wait? [Join us on Discord →](https://discord.gg/3dqQDEub9M)