What does an AI agent actually cost?
Subscription price is the wrong lens. The real monthly cost is base + token spend + ops time. Tell us your setup; we'll rank every agent in our catalog by what you'd actually pay.
Estimated monthly cost · 10 agents
Background agent that drives the Cursor editor across multi-file changes.
Base $0Ops $600Hobby tier baseline.- Free tier baseline.
- Claude CodeA78
Anthropic's terminal agent — composable, scriptable, and built around Claude's tool-use loop.
Base $0Ops $600Free tier baseline. - v0A73
Vercel's generative UI agent — design and ship React components from natural language.
Base $0Ops $600Free tier baseline. - WindsurfA78
Codeium's AI editor — Cascade agent flows alongside in-line completion and chat.
Base $0Ops $600Free tier baseline. - Replit AgentA73
Build and ship full applications from a single prompt — runs in the Replit cloud.
Base $25Ops $600Starter tier baseline. - Core tier baseline.
- Open source — base license is $0; you pay token costs and ops.
- Codex CLIB68
OpenAI’s open-source terminal agent for refactors, audits and migrations.
Base $0Tokens $60Ops $1,800Open source — base license is $0; you pay token costs and ops. - AiderB70
OSS terminal coding assistant — pair-programs, edits in place, commits as you go.
Base $0Tokens $60Ops $1,800Open source — base license is $0; you pay token costs and ops.
How we estimate
Three components:
- Base — the cheapest qualifying tier from the agent's published pricing, scaled to your seat count. For open-source agents this is $0; for sales-led agents we exclude it and flag a note.
- Tokens — for open-source agents (or BYO-key plans), the estimated monthly token spend based on your volume and the workflow's typical per-action cost (e.g. ~$1.20 per coding PR, ~$0.10 per support ticket). Managed agents include tokens in their subscription, so this is zero for them.
- Ops — engineering / admin hours per month at a loaded rate matched to your team size, times the typical hours required (more for self-hosted).
The model trades precision for honesty. Your specific token mix, dev rate, and integration depth will shift the numbers — but the relative ranking holds across reasonable parameter changes. See our full methodology.