Multi-model AI agents, billed by the token

Pick the model that fits the job — Free-1 for quick drafts, GPT-5.4 for flagship reasoning, Kimi/Grok in between. Token-based credits, no per-token surprises, switch plans anytime with prorated upgrades.

Start for free See pricing

Built for real workflows

Real model shelf

Free-1, Free-2, Kimi K2.5/K2.6, Grok 4.1/4.20, GPT-5.1 → 5.4 — picked per task, never via vendor lock-in.

Sandboxed code

Every long-running task runs in an E2B sandbox with files, terminal, and live preview.

Skills

Drop in a SKILL.md (research, design, PPT, coding) — agent picks it up automatically when the task tags match.

Bring your own MCP

Plug in any MCP server (Context7, internal tools) — managed per user with strict SSRF guards.

Sub-agents

Heavy work is fanned out via run_agent into parallel sub-agents that share a blackboard.

Resume from any step

Cancel mid-flight, switch sessions, come back later — the agent picks up where you left off.

Pricing

Free

Free

$0/month

  • 800 credits/day
  • Free-1 / Free-2 pool
  • 64k context window
  • 1 attachment / session
  • No cross-session memory
Start for free
Most popular

Plus

$7.90/month

  • 3,900 credits/day
  • 1-day rollover
  • + Kimi K2.5/K2.6, Grok-4.1 Fast
  • 100k context window
  • 1 attachment / session
  • Cross-session memory: 10 facts
Coming soon

Pro

$14.90/month

  • 7,400 credits/day
  • 2-day rollover
  • + GPT-5.1 family
  • 100k context window
  • Up to 3 attachments / session
  • Cross-session memory: 20 facts
  • Design Studio
  • Investment Research Team
Coming soon
Top tier

Max

$29.90/month

  • 14,900 credits/day
  • 2-day rollover
  • + GPT-5.2 / 5.3 / 5.4 family
  • 100k context window
  • Up to 5 attachments / session
  • Cross-session memory: 30 facts
  • Design Studio
  • Investment Research Team
  • Priority queue
Coming soon