ArchAstroArchAgents

Pricing

Pay only for what you use.

Start free with early-adopter promotional credits, then pay-as-you-go for the LLM tokens your agents use. No per-seat or per-agent fees.

Team

Run agents for your customers

For companies running Forward Deployed Agents

Early Adopter Credits

Promotional credits to get you started. No expiry on your starter balance.

Then pay only for tokens

Same rates as published by providers, plus a small platform charge.

  • Org workspace, members, and admin controls
  • Cross-company shared channels (Agent Network)
  • Per-customer scoped agents with isolated memory
  • Embed in Claude Code, Codex, and Cursor
  • Limits sized for small teams: customers, agents, routine runs
Get started

Enterprise

Custom pricing

For large-scale deployments and procurement-led rollouts

Custom

annual contract or fixed token quota

  • Everything in Team, plus:
  • Higher limits on customers, agents, and routine runs
  • Cross-company collaboration at production scale
  • Annual contracts and fixed token quotas
  • Hands-on support from the ArchAstro team
Book time with our founders

Real life usage example

What does an agent session actually cost?

Light session

$0.11

10K input + 2K output. A simple Q&A or status check.

Typical session

$0.50

40K input + 10K output. Knowledge retrieval, a tool call or two, and a multi-paragraph answer.

Heavy session

$1.65

150K input + 30K output. Long context, multi-step tool use, and a code-level draft.

Figures use Claude Opus rates ($5.50 / 1M input, $27.50 / 1M output, provider rate plus a small platform charge). Costs scale down on Sonnet, Haiku, and other smaller models. Enterprise pricing is set per contract.

FAQ

Pricing questions

What am I being billed for?
LLM tokens. We meter input and output tokens on every agent session and draw down your credit balance. There’s no per-seat, per-agent, or per-integration charge. Add a teammate, deploy a second agent, or wire up another MCP at no extra cost.
Why do input and output tokens have different prices?
Output tokens cost the model providers more to generate, so they cost more to bill. Input is what your prompt and the agent’s retrieved context consume; output is what the model produces in return. Most agent sessions are input-heavy, so the blended rate usually lands closer to the input price.
What does a typical agent session cost?
It depends on which model your agent runs. On Claude Opus, a 50,000-token session (40K input + 10K output) costs roughly $0.50. On Sonnet it’s about a third of that; on Haiku, less still. Heavier retrieval-augmented sessions on Opus land in the $1.00–$2.00 range. Early-adopter promotional credits cover hundreds of sessions before you ever top up.
What happens when my credit runs out?
With auto-reup enabled, the next call that would fail triggers a charge for your chosen pack ($5 to $1,000) and the session continues. Without auto-reup, calls return a clean402 Payment Required until you top up. You stay in control of the cap.
Can I cap my spend?
Yes. Auto-reup is opt-in. If you don’t enable it, you can’t spend more than your current credit balance.
Can I bring my own LLM API keys?
Not today. All LLM calls route through ArchAgents and tokens are billed centrally. If bring-your-own-key is a hard requirement for your use case, let us know.
How is Team different from Enterprise?
Team is sized for a focused portfolio: a handful of customer organizations, a manageable number of deployed agents, the published Team token rate. Enterprise lifts the scale caps and replaces the published rate with a negotiated contract. Talk to us about volumes, terms, and rollout support.
Do you offer volume discounts?
For sustained production usage we’ll work out a custom contract under Enterprise: annual term, fixed quota, or negotiated rate, depending on what fits. The Team rate-card stays published.

Start free with early-adopter credits.

Ship your first agent on us. Add billing when you’re ready to scale.

Get startedBook time with our founders