Enterprise
Custom
Volume pricing, contracts, security review, and support.
- Volume discounts
- Organization-wide analytics
- SSO and advanced admin controls
- SLAs and priority support
- Platform packaging with Sidekick and Visibility
Start free. Pay for what you use.
Relay bills on the tokens it routes, Sidekick on seats, and Visibility on the scope you connect. Model provider costs stay on your own accounts in every case.
Model routing that starts free and scales with traffic.
Relay sits in front of your model providers and routes each request to the best-fit model for quality, cost, and latency. Pricing is tied to routed usage and custom routing surfaces.
Pro
Your first 10M routed tokens and 3 routers are free every month. The $0.15 only applies to what you route past that.$0.15/1M tokens routed
$99/custom router/month for routing setups beyond the included routers.
- OpenAI-compatible endpoint
- Savings and usage dashboard
- Privacy-preserving routing mode
- Bring your own model keys
Model provider costs billed separately to your own accounts.
Everything included
What each plan actually gives you.
Inclusion
Pro
Enterprise
Routing
Per-request model selection
Every call is matched to a model on quality, cost, and latency rather than pinned to one.
Included routed tokens
Tokens Relay routes each month before pay-as-you-go pricing starts.
10M / month
Volume terms
Custom routers
A router trained on your own traffic and scoped to one workload.
3 free, then $99 each
Volume terms
Fallback model
If a provider fails or times out, the request completes somewhere else.
Privacy-preserving routing
Routes on request metadata so prompt contents never need to leave your control.
Integration
OpenAI-compatible endpoint
Point an existing SDK at one base URL. No client rewrite.
Bring your own model keys
Provider calls run on your accounts, so provider spend is never resold.
Shadow mode
Run a candidate beside your primary on real traffic before anything switches.
Reporting and admin
Savings and usage dashboard
What was routed, what it cost, and what the alternative would have cost.
Organization-wide analytics
Usage rolled up across every team and workspace, not just your own.
–
SSO and advanced admin controls
Directory-backed sign-in with role and policy administration.
–
SLAs and priority support
Contracted response times and a named support path.
–
Relay questions.
Relay predicts which model should handle a request based on the workload, quality target, cost, latency, and routing policy you choose.
Relay charges for the routing layer: routed tokens and custom routers. The model calls themselves still run through your own provider accounts.
A custom router is a saved routing setup for a specific workload, model pool, objective, or environment. Pro includes the base routing path; additional custom routers are $99/router/month.
Relay is focused on the routing decision. It can sit behind or alongside gateway infrastructure, but the core job is deciding which model should handle each request.
Yes. Relay is designed for long-running agent workloads where model choice, context shape, latency, and cost change across the session.
Relay can route traffic for Claude Code-style workflows when your team wants model selection, cost control, or routing policy around those sessions.
Relay evaluates request shape, context needs, past outcomes, model pool, cost objective, latency target, and your routing policy before choosing the model path.
Savings depend on traffic mix and quality requirements. The usual goal is to move requests away from unnecessarily expensive models without degrading the output your workflow needs.
Yes. Relay can learn from routing outcomes and usage patterns so recommendations improve as more company traffic flows through it.
Privacy-preserving routing limits what the router needs to inspect by using derived request features and metadata where possible instead of exposing raw payloads broadly.
Enterprise Relay can include SSO, advanced admin controls, security review support, SLAs, and priority support as part of the commercial package.
Relay is designed to keep routing overhead small, but the exact impact depends on model pool, policy, privacy mode, and where Relay sits in your request path.
Relay has a free tier, then Pro pricing based on routed tokens plus custom routers. Enterprise pricing is available for volume, controls, SLAs, and platform packaging.
Build if routing is core infrastructure your team wants to maintain. Buy if you want the routing layer, continuous model tracking, analytics, and policy controls without dedicating a research and infra team to it.
Start with a narrow workload, connect the relevant model providers, define the model pool and objective, then compare quality, latency, and cost before expanding traffic.
No. Provider keys are managed by admins and used server-side. Employees and app users do not see the keys.
A company AI app your team can actually standardize on.
Sidekick gives employees one branded AI app with company models, tools, skills, prompts, and controls. Pricing is seat-based because the value is daily team access.
Pro
7-day free trial$19/seat/month
$190 per seat/year when paid annually.
- Company-branded AI app
- Models, tools, skills, and connectors
- Workspace policy and admin controls
- Usage controls by team
- Annual billing option
Enterprise
Custom
Enterprise contracts, support, and deployment controls.
- Everything in Pro
- Larger-team pricing
- SSO and directory requirements
- Security review support
- Platform packaging with Relay and Visibility
Model provider costs billed separately to your own accounts.
Everything included
What each plan actually gives you.
Inclusion
Pro
Enterprise
The app
Company-branded AI app
One workspace your employees open by default, carrying your name rather than a vendor's.
Models, tools, and connectors
The model choices and integrations your company approves, available in one place.
Shared prompts and skills
Reusable instructions so people start from company-approved behaviour.
Controls
Workspace policy and admin
Defaults for models, tools, and behaviour, set once for everyone.
Usage controls by team
Scope which teams get which models and how much they can spend.
Server-side provider keys
Keys are held by admins and used server-side. Employees never see them.
SSO and directory requirements
Provisioning and sign-in wired to the directory you already run.
–
Commercial
Free trial
Full workspace access before any commitment.
7 days
Scoped pilot
Seat price
A seat is one active user with access to your workspace.
$19/month or $190/year
Larger-team pricing
Security review support
We work through your questionnaire and review process with you.
–
Sidekick questions.
Sidekick is a company-owned AI app for employees. It gives the team one workspace for models, prompts, skills, tools, connectors, and admin controls.
Yes. Sidekick can support company branding, workspace defaults, model choices, prompts, tools, skills, and the parts of the experience employees should see.
Yes. Admins can define workspace defaults, shared prompts, reusable skills, and approved model/tool paths so employees start from company-approved behavior.
Yes. You bring the provider keys you want Sidekick to use, and provider usage remains billed directly through those provider accounts.
No. Provider keys are managed by admins and used server-side. Employees use Sidekick without seeing the underlying keys.
You can move to Pro at $19/seat/month, annual pricing at $190/seat/year, or an enterprise package for larger teams.
No. You bring provider keys and pay provider usage directly. Sidekick is the owned app, controls, skills, and workspace layer.
A seat is one active user with access to your company Sidekick workspace.
Yes. Sidekick supports company branding, model choices, prompts, skills, connectors, and workspace controls.
AI visibility that stays current as your company changes.
Visibility shows the AI tools, accounts, workflows, usage, and spend across the company, then keeps that operating view current as teams adopt new tools.
Visibility
Custom
Scoped around connected sources, covered teams, reporting needs, and rollout depth.
- Live AI footprint map
- Source onboarding and coverage planning
- Current views for tools, accounts, workflows, usage, and spend
- Leadership-ready reporting cadence
- Platform packaging with Relay and Sidekick
Visibility pricing depends on source coverage, covered teams, reporting cadence, and rollout depth.
Everything included
What each plan actually gives you.
Inclusion
Visibility
The map
Live AI footprint map
Which AI tools, accounts, and workflows the company actually runs on.
Source onboarding
Coverage planned and connected with you rather than left as a self-serve setup.
Stays current
Observability, so the view keeps itself updated as teams adopt and drop tools.
No application code changes
The first operating map needs no engineering work from your side.
Reporting
Tools, accounts, and workflows
Current views of what is in use and who is using it.
Usage and spend
What AI costs across the company, by team and by tool.
Leadership reporting cadence
A regular readout shaped around how your leadership reviews it.
Visibility questions.
Yes. Visibility is the observability layer, so it keeps watching. The operating view updates itself as teams, tools, accounts, and usage change.
The scope depends on what sources you connect, which teams are included, and how leadership wants to monitor tools, workflows, usage, and spend.
Not for the first operating map. Deeper usage paths can be added later if your team wants more precise coverage.
Yes. Visibility is often the system of record that makes Sidekick rollout and Relay optimization easier to govern.