Integrations

Works with the AI tools your team already uses. One setting each.

Claude Code, Codex, Cursor, Copilot and your own apps reach Tempr by changing a base URL and a key. Bring keys from 34 providers, connect your MCP servers once, sign in through your identity provider, and send logs to the tools you already watch.

Start free Browse integrations

One endpoint · how each tool connects sample data

https://api.temprhq.io

Claude CodeANTHROPIC_BASE_URLplan or key

Codexmodel_provider in config.tomlplan or key

CursorOpenAI base URL overrideyour keys

GitHub CopilotBYOK provideryour keys

checkout-apibase_url in the OpenAI SDKyour keys

Everything Tempr connects to.

Each card opens its setup guide.

Coding tools 11

Point the coding agents and editors your team already runs at Tempr. Requests, costs and guardrails then cover all of them, and most guides also add the Gateway's own MCP server.

Claude CodeOn a Claude plan or an API key, with your Tempr key sent as a header.Claude plan or key Codex CLIOn a ChatGPT plan through Tempr's Codex relay, or as a custom provider.ChatGPT plan or key CursorOverride the OpenAI base URL to use any model on your allowlist.Base URL GitHub CopilotTempr as a BYOK provider, so Copilot's agent sessions can use any model.BYOK provider Claude DesktopThird-party inference mode, rolled out to every Mac and PC with a device-management profile.Managed profile ClineOne provider entry and one key for every model you've configured.Provider entry Kilo CodeA custom provider entry for every model on your allowlist.Provider entry ZedA custom provider that lists every model on your allowlist.Custom provider OpenCodeTempr as a custom provider, in one config block.Config block AiderSet OPENAI_API_BASE and your Tempr key.Environment CrushTempr as a custom provider, in one config block.Config block

AI plans 3

Keep the plans your team already pays for. The plan pays for each request, and Tempr governs it.

Claude plansPro, Max, Team and Enterprise, for Claude Code.Plan login ChatGPT plansPlus, Pro, Business, Enterprise and Edu, for Codex.Plan login Coding plansZ.AI, Kimi, MiniMax, Alibaba Cloud and more, each a provider of its own.Plan key

Apps and APIs 6

Your own apps keep their SDK. Only the base URL and the key change.

OpenAI SDKPython and TypeScript, unchanged apart from the base URL and the key./v1/chat/completions Responses APIOpenAI's newer API, for agents and for Codex./v1/responses Anthropic MessagesFor tools built on the Anthropic API, Claude Code among them./v1/messages EmbeddingsVectors for search and retrieval, from any provider's embedding models./v1/embeddings n8nRoute the OpenAI node through Tempr to call any model in a workflow.OpenAI node Management APIKeys, request logs, metrics, plan status and provider health, from your own scripts.Gateway Pro and up

MCP 6

Connect your MCP servers to Tempr once, then use them from any app, under your Tool Packs and guardrails. The Gateway has an MCP server of its own, too.

claude.ai and Claude DesktopAdd Tempr as one custom connector for every server you can reach.Connector ChatGPTThe same connector, added as a custom MCP server.Connector Claude CodeTempr's connector URL as a remote MCP server.Connector MCP catalogCurated servers, from GitHub and Postgres to Playwright, to add in one click.Catalog Your own serversAny remote MCP server, with shared credentials or a sign-in per member.Remote server The Gateway's MCP serverAsk Claude Code, Codex, claude.ai or ChatGPT about your usage, request logs and keys.Gateway Pro and up

Models and providers 3

Bring your own keys. Tempr calls each provider directly, on your key.

34 providersOpenAI, Anthropic, Google, DeepSeek, Mistral and more, on your own keys.Your keys Model catalogEvery model with its pricing, context window and capabilities.Catalog Your own modelsvLLM, SGLang or a LiteLLM proxy you run, called like any other model.Custom models

Identity 2

Sign-in and membership come from the directory you already run.

SAML and OIDCOkta, Entra ID, Google Workspace and other identity providers, with SSO required for your domain.Teams and Enterprise Directory syncSCIM adds and removes members, and mirrored groups get Tool Packs, budgets and models.Teams and Gateway Enterprise

Logs and alerts 4

Send what happens in Tempr to the tools your team already watches.

OpenTelemetryRequest logs and agent traces as spans, and the audit log as log records, over OTLP.Enterprise LangfuseRequest logs and agent traces, set up from Langfuse's own keys.Enterprise OpikRequest logs and agent traces, set up from Opik's API key and workspace.Enterprise WebhooksAlerts as JSON, each signed with an HMAC-SHA256 of its body.Every plan

Tempr's agents 4

Tempr's own agent installs into the editors your team uses.

Visual StudioTempr Chat for Visual Studio 2022 and 2026.Extension VS CodeFrom the VS Code Marketplace, or Open VSX for editors that install from it.Extension JetBrains IDEsIntelliJ IDEA, PyCharm, WebStorm, Rider and the rest.Plugin ACP editorsThe Tempr CLI as an agent in Zed, JetBrains AI Assistant, Neovim and Emacs.Agent Client Protocol

Fits the stack you have

Your AI agent, your identity provider, your log tools, your alerts.

The Gateway's MCP server

Ask an AI agent about your Gateway.

Claude Code, Codex, claude.ai and ChatGPT can read your usage and cost, request logs, keys, plan and provider health through the Gateway's own MCP server. If you allow it, an agent can make changes that only tighten, and propose the rest for you to approve in the Portal.

Sign in through the app, or use a management key
Changes that only tighten: a lower budget, fewer models, a revoked key
Anything that loosens waits for a person's approval
Gateway Pro, Scale and Enterprise

The Gateway's MCP server →

Claude Code · the Gateway's MCP server sample data

› which key spent the most this week? rein it in

↓ get_usage_summary

ci-review-bot41% of this week's spend$184.20

↓ tighten_key · propose_key_change

ci-review-bot · day budget$40 → $20tightened

checkout-api · month budget$300 → $400for your approval

Identity

Sign in the way you already do.

Members sign in through your identity provider, and directory sync adds and removes them as your directory changes. Owners can require SSO for every account on your domain.

SAML or OIDC, with Okta, Entra ID or Google Workspace
Members added and removed with your directory
Directory groups mirrored, with their own Tool Packs and budgets

Teams and SSO →

Single sign-on · acme-robotics sample data

OktaSAML · acme.io verifiedrequired

Directory syncSCIM · 6 groups mirrored42 members

priya@ added to backendfrom your directory08:02

alex@ removedleft the directory07:45

Log streaming

Your logs, in the tools you already watch.

On the Enterprise plan, an organization streams its request logs, agent traces and audit log as they happen, to up to five destinations.

OpenTelemetry over OTLP/HTTP, so any backend that takes it works
Langfuse and Opik, set up from their own keys
The audit log as signed webhook batches

Logs and streaming →

Log streaming · 3 of 5 destinations sample data

OpenTelemetry Collectorrequest logs, agent traces, audit logstreaming

Langfuserequest logs, agent tracesstreaming

Signed webhookaudit logstreaming

Alerts

Alerts where your team will see them.

Budget alerts, rule warnings and changed MCP tools arrive by email and as webhooks, each signed so you can check it came from Tempr.

Budget and rule alerts by email and webhook
mcp.tool_changed when a server changes a tool
x-tempr-webhook-signature, an HMAC-SHA256 of the body

Webhook · POST sample data

x-tempr-webhook-signature: sha256=9f2c…e41a

{ "event": "gateway.alert.triggered",
"data": { "name": "Checkout latency",
"valueText": "3,180 ms" } }

Your own apps

Two lines change. The rest of your code doesn't.

Model names carry their provider, so one key reaches every model on your allowlist.

# pip install openai
from openai import OpenAI

client = OpenAI(
    base_url="https://api.temprhq.io/v1",
    api_key="tvk_...",
)
response = client.chat.completions.create(
    model="anthropic/claude-sonnet-5.5",
    messages=[{"role": "user", "content": "Say hi"}],
)

// npm install openai
import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.temprhq.io/v1",
  apiKey: process.env.TEMPR_API_KEY,
});
const response = await client.chat.completions.create({
  model: "google/gemini-3-flash-preview",
  messages: [{ role: "user", content: "Say hi" }],
});

# any tool that reads OpenAI's environment variables
export OPENAI_API_BASE="https://api.temprhq.io/v1"
export OPENAI_API_KEY="tvk_..."
aider --model openai/anthropic/claude-sonnet-5.5

Questions

About integrations.

My tool isn't listed. Will it work?

Most likely. A tool that lets you set an OpenAI-compatible or Anthropic-compatible base URL and an API key can use Tempr: point it at https://api.temprhq.io/v1, or at https://api.temprhq.io for tools that speak the Anthropic API, and give it your Tempr key.

Do I have to change my code?

Only the base URL and the key. Requests and responses keep the shape your SDK expects, and model names carry their provider, like anthropic/claude-sonnet-5.5.

Can an AI agent change our Gateway?

Only if you allow it, with the Changes by AI agents switch. An agent connected to the Gateway's MCP server can then make changes that only tighten, like a lower budget or a revoked key. Anything that loosens, and your account's own settings, it can only propose: you approve or decline each one in the Portal.

Which integrations need a paid plan?

Coding tools, AI plans, SDKs and your own provider keys work on every Gateway plan, Free included. The management API and the Gateway's MCP server come with Pro and up, SSO and directory sync with Teams and Gateway Enterprise, and log streaming with Enterprise.

Keep going

Point one tool at Tempr today.

Start with the one your team uses most. The others take one setting each.

Start free Plan a team rollout