Skip to main content
FreeWhat would your company’s AI control layer look like?Build mine

For developers

Change the base URL.
Keep your code.

Point Cline, Claude Code, or your own apps at Hicap and reach every major model through one key. No rewrites, no new SDK.

client.ts+4 −1

1import OpenAI from "openai";
2 
3const openai = new OpenAI({
4−  apiKey: process.env.OPENAI_API_KEY,
5+  baseURL: "https://api.hicap.ai/v1",
6+  defaultHeaders: {
7+    "api-key": process.env.HICAP_API_KEY,
8+  },
9});
10 
11const res = await openai.chat.completions.create({
12  model: "gpt-5.4",
13  messages: [{ role: "user", content: prompt }],
14});
Read the docsGet your API key

Same SDK. Keep the OpenAI client you already ship, in any language.

Same API calls. Chat completions, embeddings, streaming, and tool calls: unchanged.

Same responses. The response format is identical, so nothing downstream needs to know.

By prompt

Let your coding agent
do the migration.

Paste this instruction into Claude Code, Cursor, Cline, or another coding agent. It reads the setup guide on docs.hicap.ai and helps update your project.

  • Finds every client

    Searches for new OpenAI, createOpenAI, OpenAIClient, and helper factories.

  • Moves the key

    Adds HICAP_API_KEY to your env files and drops the old Authorization headers.

  • Leaves your calls alone

    chat.completions, embeddings, and every other call site stay untouched.

Setup with your coding agent

Read https://docs.hicap.ai/onboarding-prompts/configure-client and help me set this up.

Review full prompt on docs.hicap.ai

By setup guide

Exact steps
for your stack.

Config examples and the fastest path from API key to working integration, tool by tool.

Open the integrations libraryOr browse the documentation

Verify

Confirm it
in one call.

Drop in your key and check the endpoint is live before you wire up your tools.

  1. 01

    Export your key.

    Set HICAP_API_KEY in your shell from the key you created in the dashboard.

  2. 02

    Send one request.

    Any model string works. Try a second provider by changing only that string.

  3. 03

    Find it in the log.

    The request shows up with the model you asked for, the model that answered, and its cost.

1curl https://api.hicap.ai/v1/chat/completions \
2  -H "api-key: $HICAP_API_KEY" \
3  -H "Content-Type: application/json" \
4  -d '{
5    "model": "gpt-5.5",
6    "messages": [
7      { "role": "user", "content": "Hello" }
8    ]
9  }'

Solutions

Bring the whole team.
On one key and one bill.

Volume pricing, shared billing, and dedicated support for your whole engineering team.

Talk to our engineers

Reserved capacity routing

Cut your AI spend by up to 30%.

Hicap routes your traffic onto reserved capacity at a fixed price and overflows to on-demand when demand spikes. Same models, same code, a smaller bill.

  1. 01

    Reserved capacity first

    Your baseline runs on capacity bought at a fixed price.

  2. 02

    On-demand for the burst

    Spikes spill over automatically. Nothing queues, nothing drops.

  3. 03

    One bill, up to 30% lower

    Same models, same code. Only the invoice changes.

Get startedSee what you’d save