Pricing
Start free and scale as you grow.
What you get
Agent Identity
Every agent acts under a persistent identity.
Access governed by your identity providerShort-lived, task-scoped credentialsEvery decision attributed to a name
MCP Gateway
One governed path to every server.
One official registry of MCP serversShadow MCP discovered and promotedEvery tool call logged to your OTEL destination
Agent Security
Every prompt and tool call checked.
Policy enforced on every actionPrompt injection, and data exfiltration protectionRunaway sessions quarantined
What you pay
Pay as you go
No platform fee. Pay only for what you use
Usage
$0.35 per million tokens of agent session storage
$0.99 per million tokens scanned for risk
$20 per GiB of MCP gateway egress
Enterprise
Every AI transformation is different
Pay as you go, plus
Annual contract on tokens under management
Generous caps for session storage, MCP gateway throughput, and risk analysis
SSO and full audit trail
Early feature access
Legal and security reviews
Support
SLA-backed support
Dedicated FDE support
Concierge onboarding
Dedicated support channel
Frequently AskedQuestions
What does Enterprise add over pay as you go?
Both plans carry the same control plane: the same connectors, the same OAuth, provisioning and RBAC controls, and the same logs, tracing and analytics. Enterprise adds an annual contract priced on tokens under management, SSO and a full audit trail, SLA-backed support with a dedicated forward-deployed engineer and support channel, concierge onboarding, early feature access, and legal and security reviews.
How does pay-as-you-go billing work?
There is no platform fee. You're billed each month for the tokens of agent session storage you use ($0.35 per million), the tokens each risk scanner processes ($0.99 per million), and MCP gateway egress ($20 per GiB). Other inference is passed through at provider cost.
What are tokens under management?
Tokens under management is the volume of agent traffic the platform observes each billing cycle, measured in tokens. It counts the prompt and response tokens in the agent sessions Speakeasy is instrumented on, the content written to a prompt cache for the first time, and MCP tool calls. Cache reads don't count, because that content was already counted once when it was written.
How does Speakeasy count a token?
Speakeasy meters in its own unit, the Speakeasy token, and applies it to the content it processes from your agent sessions: user prompts, assistant messages, and tool names, inputs and outputs. Model providers tokenize the same text differently, so identical sessions report different totals depending on the model behind them, and provider counts fold in content Speakeasy never touches, such as reasoning, compaction, and the small background calls that write chat titles. Counting the same way every time keeps the same sessions at the same cost whichever models and AI tools your teams run.
What counts as agent session storage?
Agent session storage meters the stored-message workload from the sessions Speakeasy captures, measured in tokens rather than provider input and output tokens. The usage explorer reports it by UTC day, and breaks it down by project, model and provider, or by the divisions, departments and roles behind it.
Does the inference Speakeasy runs count toward my bill?
It's billed separately. Playground completions, assistant runs and risk-policy judges are inference the platform runs on your behalf, so they're passed through at provider cost and excluded from tokens under management. Inference caps limit what that can cost each calendar month, set per platform-managed key. Organizations that bring their own model provider keys manage those budgets on the provider side instead.
Is there a free trial?
Yes. Pay as you go starts with a trial, and the billing page lists the rates that apply once it converts. Add a payment method before the trial ends and the organization moves onto pay as you go at those rates.