|
| 1 | +--- |
| 2 | +title: Pydantic AI Gateway |
| 3 | +status: new |
| 4 | +--- |
| 5 | + |
| 6 | +# Pydantic AI Gateway |
| 7 | + |
| 8 | +**[Pydantic AI Gateway](https://pydantic.dev/gateway)** (PAIG) is a unified interface for accessing multiple AI providers with a single key. Features include built-in OpenTelemetry observability, real-time cost monitoring, failover management, and native integration with the other tools in the [Pydantic stack](https://pydantic.dev/). |
| 9 | + |
| 10 | +!!! note "Free while in Beta" |
| 11 | + The Pydantic AI Gateway is currently in Beta. You can bring your own key (BYOK) or buy inference through the Gateway (we will eat the card fee for now). |
| 12 | + |
| 13 | +Sign up at [gateway.pydantic.dev](https://gateway.pydantic.dev/). |
| 14 | + |
| 15 | +!!! question "Questions?" |
| 16 | + For questions and feedback, contact us on [Slack](https://logfire.pydantic.dev/docs/join-slack/). |
| 17 | + |
| 18 | +## Documentation Integration |
| 19 | + |
| 20 | +To help you get started with [Pydantic AI Gateway](https://gateway.pydantic.dev), some code examples on the Pydantic AI documentation include a "Via Pydantic AI Gateway" tab, alongside a "Direct to Provider API" tab with the standard Pydantic AI model string. The main difference between them is that when using Gateway, model strings use the `gateway/` prefix. |
| 21 | + |
| 22 | +## Key features |
| 23 | + |
| 24 | +- **API key management**: access multiple LLM providers with a single Gateway key. |
| 25 | +- **Cost Limits**: set spending limits at project, user, and API key levels with daily, weekly, and monthly caps. |
| 26 | +- **BYOK and managed providers:** Bring your own API keys (BYOK) from LLM providers, or pay for inference directly through the platform. |
| 27 | +- **Multi-provider support:** Access models from OpenAI, Anthropic, Google Vertex, Groq, and AWS Bedrock. _More providers coming soon_. |
| 28 | +- **Backend observability:** Log every request through [Pydantic Logfire](https://pydantic.dev/logfire) or any OpenTelemetry backend (_coming soon_). |
| 29 | +- **Zero translation**: Unlike traditional AI gateways that translate everything to one common schema, PAIG allows requests to flow through directly in each provider's native format. This gives you immediate access to the new model features as soon as they are released. |
| 30 | +- **Open source with self-hosting**: PAIG's core is [open source](https://github.com/pydantic/pydantic-ai-gateway/) (under [AGPL-3.0](https://www.gnu.org/licenses/agpl-3.0.en.html)), allowing self-hosting with file-based configuration, instead of using the managed service. |
| 31 | +- **Enterprise ready**: Includes SSO (with OIDC support), granular permissions, and flexible deployment options. Deploy to your Cloudflare account, or run on-premises with our [consulting support](https://pydantic.dev/contact). |
| 32 | + |
| 33 | +```python {title="hello_world.py"} |
| 34 | +from pydantic_ai import Agent |
| 35 | + |
| 36 | +agent = Agent('gateway/openai:gpt-5') |
| 37 | + |
| 38 | +result = agent.run_sync('Where does "hello world" come from?') |
| 39 | +print(result.output) |
| 40 | +""" |
| 41 | +The first known use of "hello, world" was in a 1974 textbook about the C programming language. |
| 42 | +""" |
| 43 | +``` |
| 44 | +# Quick Start |
| 45 | + |
| 46 | +This section contains instructions on how to set up your account and run your app with Pydantic AI Gateway credentials. |
| 47 | + |
| 48 | +## Create an account |
| 49 | + |
| 50 | +Using your GitHub or Google account, sign in at [gateway.pydantic.dev](https://gateway.pydantic.dev). |
| 51 | +Choose a name for your organization (or accept the default). You will automatically be assigned the Admin role. |
| 52 | + |
| 53 | +A default project will be created for you. You can choose to use it, or create a new one on the [Projects](https://gateway.pydantic.dev/admin/projects) page. |
| 54 | + |
| 55 | +## Add **Providers** |
| 56 | +There are two ways to use Providers in the Pydantic AI Gateway: you can bring your own key (BYOK) or buy inference through the platform. |
| 57 | + |
| 58 | +### Bringing your own API key (BYOK) |
| 59 | + |
| 60 | +On the [Providers](https://gateway.pydantic.dev/admin/providers) page, fill in the form to add a provider. Paste your API key into the form under Credentials, and make sure to **select the Project that will be associated to this provider**. It is possible to add multiple keys from the same provider. |
| 61 | + |
| 62 | +### Use Built-in Providers |
| 63 | +Go to the Billing page, add a payment method, and purchase $15 in credits to activate built-in providers. This gives you single-key access to all available models from OpenAI, Anthropic, Google Vertex, AWS Bedrock, and Groq. |
| 64 | + |
| 65 | +## Grant access to your team |
| 66 | +On the [Users](https://gateway.pydantic.dev/admin/users) page, create an invitation and share the URL with your team to allow them to access the project. |
| 67 | + |
| 68 | +## Create Gateway project keys |
| 69 | +On the Keys page, Admins can create project keys which are not affected by spending limits. Users can only create personal keys, that will inherit spending caps from both User and Project levels, whichever is more restrictive. |
| 70 | + |
| 71 | +# Usage |
| 72 | +After setting up your account with the instructions above, you will be able to make an AI model request with the Pydantic AI Gateway. |
| 73 | +The code snippets below show how you can use PAIG with different frameworks and SDKs. |
| 74 | +You can add `gateway/` as prefix on every known provider that |
| 75 | + |
| 76 | +To use different models, change the model string `gateway/<api_format>:<model_name>` to other models offered by the supported providers. |
| 77 | + |
| 78 | +Examples of providers and models that can be used are: |
| 79 | + |
| 80 | +| **Provider** | **API Format** | **Example Model** | |
| 81 | +| --- |-----------------|------------------------------------------| |
| 82 | +| OpenAI | `openai` | `gateway/openai:gpt-5` | |
| 83 | +| Anthropic | `anthropic` | `gateway/anthropic:claude-sonnet-4-5` | |
| 84 | +| Google Vertex | `google-vertex` | `gateway/google-vertex:gemini-2.5-flash` | |
| 85 | +| Groq | `groq` | `gateway/groq:openai/gpt-oss-120b` | |
| 86 | +| AWS Bedrock | `bedrock` | `gateway/bedrock:amazon.nova-micro-v1:0` | |
| 87 | + |
| 88 | +## Pydantic AI |
| 89 | +Before you start, make sure you are on version 1.16 or later of `pydantic-ai`. To update to the latest version run: |
| 90 | + |
| 91 | +=== "uv" |
| 92 | + |
| 93 | + ```bash |
| 94 | + uv sync -P pydantic-ai |
| 95 | + ``` |
| 96 | + |
| 97 | +=== "pip" |
| 98 | + |
| 99 | + ```bash |
| 100 | + pip install -U pydantic-ai |
| 101 | + ``` |
| 102 | + |
| 103 | +Set the `PYDANTIC_AI_GATEWAY_API_KEY` environment variable to your Gateway API key: |
| 104 | + |
| 105 | +```bash |
| 106 | +export PYDANTIC_AI_GATEWAY_API_KEY="YOUR_PAIG_TOKEN" |
| 107 | +``` |
| 108 | + |
| 109 | +You can access multiple models with the same API key, as shown in the code snippet below. |
| 110 | + |
| 111 | +```python {title="hello_world.py"} |
| 112 | +from pydantic_ai import Agent |
| 113 | + |
| 114 | +agent = Agent('gateway/openai:gpt-5') |
| 115 | + |
| 116 | +result = agent.run_sync('Where does "hello world" come from?') |
| 117 | +print(result.output) |
| 118 | +""" |
| 119 | +The first known use of "hello, world" was in a 1974 textbook about the C programming language. |
| 120 | +""" |
| 121 | +``` |
| 122 | + |
| 123 | + |
| 124 | +## Claude Code |
| 125 | +Before you start, log out of Claude Code using `/logout`. |
| 126 | + |
| 127 | +Set your gateway credentials as environment variables: |
| 128 | + |
| 129 | +```bash |
| 130 | +export ANTHROPIC_BASE_URL="https://gateway.pydantic.dev/proxy/anthropic" |
| 131 | +export ANTHROPIC_AUTH_TOKEN="YOUR_PAIG_TOKEN" |
| 132 | +``` |
| 133 | + |
| 134 | +Replace `YOUR_PAIG_TOKEN` with the API key from the Keys page. |
| 135 | + |
| 136 | +Launch Claude Code by typing `claude`. All requests will now route through the Pydantic AI Gateway. |
| 137 | + |
| 138 | +## SDKs |
| 139 | + |
| 140 | +### OpenAI SDK |
| 141 | + |
| 142 | +```python {title="openai_sdk.py" test="skip"} |
| 143 | +import openai |
| 144 | + |
| 145 | +client = openai.Client( |
| 146 | + base_url='https://gateway.pydantic.dev/proxy/chat/', |
| 147 | + api_key='paig_...', |
| 148 | +) |
| 149 | + |
| 150 | +response = client.chat.completions.create( |
| 151 | + model='gpt-5', |
| 152 | + messages=[{'role': 'user', 'content': 'Hello world'}], |
| 153 | +) |
| 154 | +print(response.choices[0].message.content) |
| 155 | +#> Hello user |
| 156 | +``` |
| 157 | + |
| 158 | +### Anthropic SDK |
| 159 | + |
| 160 | +```python {title="anthropic_sdk.py" test="skip"} |
| 161 | +import anthropic |
| 162 | + |
| 163 | +client = anthropic.Anthropic( |
| 164 | + base_url='https://gateway.pydantic.dev/proxy/anthropic/', |
| 165 | + auth_token='paig_...', |
| 166 | +) |
| 167 | + |
| 168 | +response = client.messages.create( |
| 169 | + max_tokens=1000, |
| 170 | + model='claude-sonnet-4-5', |
| 171 | + messages=[{'role': 'user', 'content': 'Hello world'}], |
| 172 | +) |
| 173 | +print(response.content[0].text) |
| 174 | +#> Hello user |
| 175 | +``` |
0 commit comments