Pricing
Pricing that pays for itself.
LLM usage billed at cost, plus a flat $10 per 100K requests. Cohesor typically cuts agent spend 60–90% — so the gateway more than covers its own cost from day one.
Pay as you go
For teams running and building with agents.
$10 / 100K requests
platform fee — LLM usage billed at cost
Get started- Token compression & smart routing
- Unlimited gateway keys
- MCP tools (Slack, GitHub, Google)
- Per-user budgets & domain join
- Full history & billing reports
- Email support
Enterprise
For organizations at scale.
Custom
volume pricing & terms
Contact sales- Everything in Pay as you go, plus:
- Volume discounts on the platform fee
- SSO & SCIM provisioning
- Custom & self-hosted models
- SLA, dedicated support, security review & DPA
You pay LLM usage at provider cost plus a flat $10 per 100K requests platform fee. You keep the compression & routing savings.
Savings calculator
Estimate what Cohesor saves you.
Computed live from real per-token model rates — flagship $3/$15 vs. economy $0.20/$0.60 per 1M, with ~50% prompt compression.
Simple requests route to the economy model.
Estimated monthly spend
Without Cohesor
$0
With Cohesor
$0
You save
0%
$0 / month
≈ $0 / year
Estimate only. Actual savings depend on your prompts, model mix and routing threshold. Compression assumed at 50% of prompt tokens; the "With Cohesor" figure includes the $10 / 100K requests platform fee.
FAQ
Questions, answered.
A simple platform fee of $10 per 100,000 requests. LLM usage is billed at cost — and because we cut your underlying spend so sharply, your all-in bill still drops dramatically.
No. Point your agent at Cohesor with one environment variable. Claude Code, Codex and any Anthropic/OpenAI-compatible client work unchanged.
Compression preserves meaning, not every byte. Routing only sends genuinely simple work to cheaper models — and you control the sensitivity, or pin a single model in fixed mode.
Yes. Use our curated catalog or pass any model slug directly. Enterprise plans support custom and self-hosted models.
Provider keys are encrypted at rest and never returned to the browser; gateway keys are hashed; MCP OAuth tokens stay client-side. See our security overview.