Integrations · Direct
When to use Direct
Your backend calls Pikeway with your own key. Every call is billed to your balance, and your users only ever talk to your product.
How it works
You create one or more keys in the console and keep them on your server. Your code calls /v1/run (or /gw/<slug>/…) server to server. Pikeway never sees your end users; to Pikeway, every call comes from your account.
Who pays
You. Calls are charged to your prepaid balance at list price. How you recover that cost (subscription, usage fee, free) is up to you.
Who owns the user
You. Users don't need a Pikeway account and never see Pikeway, unless you tell them.
Good fit
- Your own agents and automations: research bots, trading signals, data pipelines, CI jobs.
- A backend feature inside your product, such as enriching a profile or fetching a token price, where the API is an implementation detail.
- Prototypes and hackathons: one key, one env var, first call in minutes.
- Teams that run many agents and want a separate budget for each, without building billing.
Pros and cons
Pros
- Least work: no OAuth, no proxy, nothing to operate.
- Invisible to users: no extra sign-up or consent screen.
- Fine-grained guardrails per key: spending limit, IP allowlist, and budget-limited agent keys under Console → Team.
- One balance and one invoice trail for everything your company runs.
Cons
- You carry the cost of every call, including abuse by your own users. You need your own rate limits or quotas per user.
- The key must stay on a server. Browser or mobile use needs a separate browser key restricted to your sites.
- No per-user attribution in Pikeway: all calls belong to your account. Use one key per tenant if you need that split, or switch to a proxy.
Compared with the other approaches
| Direct | Sign in with Pikeway | Your own proxy | |
|---|---|---|---|
| Who pays Pikeway | You, from your balance | Each user, from their own balance | You; you bill your users however you like |
| Owns the user relationship | You. Users never see Pikeway | Shared. Users have a Pikeway account and approve your app | You. Users never see Pikeway |
| User needs a Pikeway account | No | Yes | No |
| Your own pricing | Bundle the cost into your product | No: users pay Pikeway list prices | Yes: markup, credits, plans |
| Spend controls | Per-key limits, agent keys, IP allowlist | User picks a limit at consent; revocable any time | Your own rules, plus key limits as a backstop |
| What you run | Nothing extra: call from your backend | An OAuth client and token storage | A proxy service you operate |
| Time to first call | Minutes | About an hour | A day or more |
Users should pay instead?
Sign in with Pikeway moves the cost to each user's own balance.
Need your own pricing?
A proxy adds your auth, prices and endpoint allow-list in front of the same key.
Direct is also the right choice when the "user" is you: connecting your own Claude, Cursor or Codex to Pikeway. See the MCP quickstart.