Configure JWT-based authentication for your organization in Portkey
This feature is available only on the Enterprise Plan of Portkey.
Portkey supports JWT-based authentication in addition to API key authentication. Clients send a JWT in x-portkey-api-key or Authorization: Bearer; the token is validated against JWKS configured on the organisation.Portkey Cloud validates JWTs on the control plane. Hybrid and air-gapped deployments can optionally enable gateway-local JWT authentication so the AI Gateway validates tokens locally instead of delegating every request to the control plane.
Validate JWT Token (Guardrail)
Optionally validate JWTs inside a config’s hook pipeline using the JWT guardrail plugin. This is separate from gateway-local JWT auth, which replaces API key validation at the gateway.
Your JWT payload must contain the following claims:
Claim Key
Description
portkey_oid / organisation_id
Unique identifier for the organization.
portkey_workspace / workspace_slug
Identifier for the workspace.
scope / scopes
Permissions granted by the token.
exp
Expiration time (as a UNIX timestamp, in seconds).
exp is mandatory. Tokens without exp or with expired exp are rejected. A small amount of clock skew is tolerated.
iat and/or nbf are recommended but optional.
On Portkey Cloud, portkey_oid and portkey_workspace are required. In gateway-local JWT auth, org and workspace can be resolved from deployment configuration when you use a standard IdP token - see Integration modes.
If your JWT, JWKS, and claims are correct, the request will authenticate and succeed.
If you prefer Python for signing, you can generate the RSA key pair using your preferred method, ensure the public key is present in your JWKS with a matching kid, and use a library like PyJWT to sign with RS256 while setting the header { "alg": "RS256", "typ": "JWT", "kid": "<your-kid>" }.
The client sends the JWT in x-portkey-api-key or Authorization: Bearer:
x-portkey-api-key: <JWT_TOKEN>
If x-portkey-api-key is absent, Authorization is used.
The gateway or control plane validates the JWT:
Verifies the signature using the organisation JWKS.
Checks token expiry.
Ensures required claims are present.
If valid, the request is authenticated and user details are extracted for authorization and logging.
If invalid, the request is rejected. Common responses include 401 Unauthorized (invalid or expired token), 403 Forbidden (scope or workspace not allowed), 412 (usage limit exceeded), and 429 (rate limit exceeded).
On hybrid and air-gapped gateways with JWT_ENABLED=ON, validation runs locally on the gateway. Otherwise, JWT-shaped tokens are validated by the control plane.
Once the JWT is validated, the server checks for the required scope. Scopes can be provided in the JWT as either a single string or an array of strings using the scope or scopes claim.
Scopes can also be prefixed with portkey. (e.g., portkey.completions.write).
JWT tokens with appropriate scopes function identically to workspace API keys, providing access to workspace-specific operations. They cannot be used as organization API keys, which have broader administrative permissions across all workspaces.
You can embed a default config directly in the JWT payload using the defaults claim. This works the same way as attaching default configs to API keys, but lets you control it per-token - useful when different users or services need different routing rules.The defaults object supports the following fields:
Field
Description
config_id
The ID of the config to apply as the default for all requests made with this token.
metadata
A JSON object with custom metadata to attach to all requests.
When a JWT with a defaults.config_id is used, Portkey validates that the config belongs to the same organization before applying it. If the config is not found or doesn’t belong to the org, the default is ignored.
This follows the same precedence rules as API key default configs - if a user explicitly passes a config ID in their request headers, it will override the JWT default unless config override is disabled at the workspace level.
Once you have a valid JWT token, you can use it to authenticate your API calls to Portkey. Below are examples showing how to use JWT authentication with different SDKs.
NodeJS
Python
cURL
OpenAI Python SDK
OpenAI NodeJS SDK
Install the Portkey SDK with npm
npm install portkey-ai
import Portkey from 'portkey-ai';const client = new Portkey({ apiKey: '<JWT_TOKEN>', // Use JWT token instead of API key provider: '@<YOUR_PROVIDER_SLUG>'});async function main() { const response = await client.chat.completions.create({ messages: [{ role: "user", content: "Hello, how are you today?" }], model: "your-model-id", }); console.log(response.choices[0].message.content);}main();
Install the Portkey SDK with pip
pip install portkey-ai
from portkey_ai import Portkeyclient = Portkey( api_key = "<JWT_TOKEN>", # Use JWT token instead of API key, provider = "@<YOUR_PROVIDER_SLUG>")response = client.chat.completions.create( model="your-model-id", messages=[ {"role": "system", "content": "You are a helpful assistant."}, {"role": "user", "content": "Hello!"} ])print(response.choices[0].message)
Wrong algorithm: Only RS256 is accepted; HS256 or others will fail.
Missing or mismatched kid: Your JWT header must include a kid that matches a key in the JWKS.
Incorrect header usage: Send the raw JWT in x-portkey-api-key, or use Authorization: Bearer <JWT>. Do not prefix the token with Bearer in x-portkey-api-key.
Expired or missing exp: The exp claim is required and must be in the future. Allow for small clock skew.
Private vs Public key mix-up: Your JWKS must contain only the public key parameters. The private key is used only for signing; never paste it into the JWKS JSON.
Wrong org/workspace identifiers: portkey_oid (or organisation_id) and portkey_workspace (or workspace_slug) must correspond to valid identifiers in your Portkey tenant.
Scopes missing for the API you call: E.g., chat completions needs completions.write.
Unreachable JWKS URL: If using a URL, it must be publicly reachable by Portkey. For static JSON, ensure the pasted JSON is valid and includes keys: [...].
All Invalid JWT errors are logged in the Audit Logs.
Sample error message:
"jwtToken": "ey****bt","error": "Error Verifying JWT token: Signing Key Not Found",
Gateway-local JWT authentication is available only on hybrid and air-gapped deployments.It requires gateway 2.5.0 or higher (Backend v1.13.0 or higher for air-gapped) with JWT_ENABLED=ON.
When enabled, your AI Gateway validates JWTs locally instead of sending every request to the control plane for authentication. This reduces latency and keeps authentication working even when the control plane is temporarily unreachable.If JWT_ENABLED is not ON, JWT-shaped tokens are validated by the control plane instead.
You can add Portkey-specific claims to your tokens (portkey_oid, portkey_workspace, usage_limits, rate_limits, defaults, and others) and use the full JWT feature set described in this guide.
Use this when tokens come from Okta, Auth0, Entra, Cognito, or another IdP whose token schema you cannot change. Portkey handles org, workspace, and budget configuration on the server side. If your IdP cannot include Portkey scopes in tokens, set JWT_LOCAL_AUTH_DEFAULT_SCOPES on the gateway instead.
What you need
How to configure it
Organisation
Set ORGANISATIONS_TO_SYNC to a single org UUID when the token does not include portkey_oid or organisation_id.
Workspace
Map users to workspaces with deployment settings, restrict the deployment to specific workspaces, or set an org-level default workspace. See Workspace resolution.
Scopes
Register gateway scopes (e.g. completions.write) in your IdP, or set JWT_LOCAL_AUTH_DEFAULT_SCOPES on the gateway when IdP scope changes are not possible. Both completions.write and portkey.completions.write are accepted.
Usage and rate limits
Configure limits on the workspace, or use workspace policies keyed off request metadata.
User attribution in logs
The gateway uses email_id, sub, or uid from the token for per-user logging and metrics.
JWKS
Point your org’s JWKS URL to your IdP’s discovery endpoint (e.g. https://<tenant>/.well-known/jwks.json) in Admin Settings → Organisation → Authentication.
Set to ON to enable gateway-local JWT auth. Any other value uses control plane validation.
ORGANISATIONS_TO_SYNC
Yes
Comma-separated org UUIDs to sync from the control plane. Also restricts which orgs this gateway accepts.
PORTKEY_CLIENT_AUTH
Yes
Service auth token for the gateway to sync org, workspace, and deployment settings from the control plane.
JWT_LOCAL_AUTH_DEFAULT_SCOPES
No
Comma-separated gateway scopes to apply when the IdP token does not include scope or scopes. Use this when you cannot add Portkey scopes to tokens in your IdP.
If the token omits scope / scopes, set JWT_LOCAL_AUTH_DEFAULT_SCOPES on the gateway (e.g. completions.write,mcp.invoke).In Mode B, the gateway resolves the org from ORGANISATIONS_TO_SYNC and the workspace from your deployment configuration.
Scopes can be provided as a space-separated string (scope) or an array (scopes) in the token, or via JWT_LOCAL_AUTH_DEFAULT_SCOPES on the gateway. A portkey. prefix is optional.
Scope
Used for
completions.read
Read completions
completions.write
Chat completions, embeddings, and most AI routes
logs.read
Read logs
logs.write
Write logs
logs.export
Export logs
prompts.render
Render prompts
virtual_keys.list
List virtual keys
mcp.invoke
MCP invocation
agents.invoke
Agent invocation
guardrails.invoke
Guardrail invocation
feedbacks.read
Read feedbacks
feedbacks.write
Write feedbacks
Each API route requires a matching scope. For example, chat completions requires completions.write.
Workspace is determined after the token is verified, using token claims first and then deployment settings.From the token:
portkey_workspace
workspace_slug
From deployment settings (when the token has no workspace claim):
Setting
Effect
User allowlist
Restrict authentication to specific user IDs (sub values)
User-to-workspace mapping
Map each user ID to a workspace slug
Workspace allowlist
When enabled, only listed workspaces are allowed; the first listed workspace is used as the default
Org default workspace
Used when no workspace is resolved from the token or deployment settings
Contact your Portkey account team to configure deployment-level JWT settings such as user allowlists, user-to-workspace mappings, and workspace allowlists.
JWT-authenticated requests are subject to the same limit and policy checks as API key requests. Limits can come from three places:
Source
Description
Token claims
usage_limits and rate_limits embedded in the JWT (Mode A)
Workspace settings
Limits configured on the resolved workspace
Workspace policies
Conditional budget policies based on request context
Each unique JWT is tracked separately for token-level limits. Workspace limits and policies apply on top of any token-level limits - all must pass before the request proceeds.Guidance by mode:
Mode A: Put per-user or per-session budgets in the JWT usage_limits claim. Use workspace limits and policies for shared team budgets.
Mode B: Configure all budgets and rate limits on the workspace or as workspace policies. For per-user splits within a shared workspace, use policies keyed off user metadata. For hard per-user isolation, map each user to a dedicated workspace.
Validated JWTs are cached until they expire to reduce validation overhead. On gateway-local auth, org and workspace settings are also cached locally and refreshed during control plane sync.If you rotate signing keys:
Publish the new public key in JWKS with a new kid.
Start issuing tokens signed by the new private key (with the new kid).
Old tokens remain valid until their exp is reached.