Create API key

Create a new API key for user authentication and authorization.

Use this endpoint when users sign up, upgrade subscription tiers, or need additional keys. Keys are cryptographically secure and unique to the specified API namespace.

Important: The key is returned only once. Store it immediately and provide it to your user, as it cannot be retrieved later.

Common use cases:

Generate keys for new user registrations
Create additional keys for different applications
Issue keys with specific permissions or limits

Required Permissions

Your credential needs one of:

api.*.create_key (create keys in any API)
api.<api_id>.create_key (create keys in specific API)
unkey:v1:<workspace_id>:keyspaces/*#create_key (create keys in any keyspace)
unkey:v1:<workspace_id>:keyspaces/<keyspace_id>#create_key (create keys in a specific keyspace)

1 min read

post/v2/keys.createKey

Request example

package main

import(
	"context"
	"os"
	unkey "github.com/unkeyed/sdks/api/go/v2"
	"github.com/unkeyed/sdks/api/go/v2/models/components"
	"log"
)

func main() {
    ctx := context.Background()

    s := unkey.New(
        unkey.WithSecurity(os.Getenv("UNKEY_ROOT_KEY")),
    )

    res, err := s.Keys.CreateKey(ctx, components.V2KeysCreateKeyRequestBody{
        APIID: "api_1234abcd",
        Prefix: unkey.Pointer("prod"),
        Name: unkey.Pointer("Payment Service Production Key"),
        ByteLength: unkey.Pointer[int64](24),
        ExternalID: unkey.Pointer("user_1234abcd"),
        Meta: map[string]any{
            "plan": "enterprise",
            "featureFlags": map[string]any{
                "betaAccess": true,
                "concurrentConnections": 10,
            },
            "customerName": "Acme Corp",
            "billing": map[string]any{
                "tier": "premium",
                "renewal": "2024-12-31",
            },
        },
        Roles: []string{
            "api_admin",
            "billing_reader",
        },
        Permissions: []string{
            "documents.read",
            "documents.write",
            "settings.view",
        },
        Expires: unkey.Pointer[int64](1704067200000),
        Credits: &components.KeyCreditsData{
            Remaining: unkey.Pointer[int64](1000),
            Refill: &components.KeyCreditsRefill{
                Interval: components.KeyCreditsRefillIntervalDaily,
                Amount: 1000,
                RefillDay: unkey.Pointer[int64](15),
            },
        },
        Ratelimits: []components.RatelimitRequest{
            components.RatelimitRequest{
                Name: "requests",
                Limit: 100,
                Duration: 60000,
                AutoApply: unkey.Pointer(true),
            },
            components.RatelimitRequest{
                Name: "heavy_operations",
                Limit: 10,
                Duration: 3600000,
            },
        },
    })
    if err != nil {
        log.Fatal(err)
    }
    if res.V2KeysCreateKeyResponseBody != nil {
        // handle response
    }
}

Response

{
  "meta": {
    "requestId": "req_123"
  },
  "data": {
    "keyId": "key_2cGKbMxRyIzhCxo1Idjz8q",
    "key": "prod_2cGKbMxRjIzhCxo1IdjH3arELti7Sdyc8w6XYbvtcyuBowPT"
  }
}

post/v2/keys.createKey

Authorization

Authorizationstringheaderrequired#

Unkey uses bearer tokens for authentication. Public integrations use root keys, while the dashboard proxy uses short-lived JWTs.
To authenticate, include the token in the Authorization header of each request:

Authorization: Bearer unkey_123

Root keys have specific permissions attached to them, controlling what operations they can perform. Legacy permissions use tuple strings like api.*.create_key; resource permissions use Unkey Resource Names plus actions, like unkey:v1:ws_123:keyspaces/*#create_key.
Security best practices:

Keep root keys secure and never expose them in client-side code
Use different root keys for different environments
Rotate keys periodically, especially after team member departures
Create keys with minimal necessary permissions following least privilege principle
Monitor key usage with audit logs.

Body

application/json

apiIdstringrequired#

The API namespace this key belongs to.
Keys from different APIs cannot access each other.

Length: 3–255Pattern: ^[a-zA-Z0-9_]+$

prefixstring#

Adds a visual identifier to the beginning of the generated key for easier recognition in logs and dashboards.
The prefix becomes part of the actual key string (e.g., prod_xxxxxxxxx).
Avoid using sensitive information in prefixes as they may appear in logs and error messages.

Length: 1–16Pattern: ^[a-zA-Z0-9_]+$

namestring#

Sets a human-readable identifier for internal organization and dashboard display.
Never exposed to end users, only visible in management interfaces and API responses.
Avoid generic names like "API Key" when managing multiple keys for the same user or service.

Length: 1–255

byteLengthinteger#

Controls the cryptographic strength of the generated key in bytes.
Higher values increase security but result in longer keys that may be more annoying to handle.
The default 16 bytes provides 2^128 possible combinations, sufficient for most applications.
Consider 32 bytes for highly sensitive APIs, but avoid values above 64 bytes unless specifically required.

Default: 16

Range: 16–255

externalIdstring#

Links this key to a user or entity in your system using your own identifier.
Returned during verification to identify the key owner without additional database lookups.
Essential for user-specific analytics, billing, and multi-tenant key management.
Use your primary user ID, organization ID, or tenant ID for best results.
Accepts letters, numbers, underscores, dots, and hyphens for flexible identifier formats.

Length: 1–255Pattern: ^[a-zA-Z0-9_.-]+$

metaobject#

Stores arbitrary JSON metadata returned during key verification for contextual information.
Eliminates additional database lookups during verification, improving performance for stateless services.
Avoid storing sensitive data here as it's returned in verification responses.
Large metadata objects increase verification latency and should stay under 10KB total size.

rolesstring[]#

Assigns existing roles to this key for permission management through role-based access control.
Roles must already exist in your workspace before assignment.
During verification, all permissions from assigned roles are checked against requested permissions.
Roles provide a convenient way to group permissions and apply consistent access patterns across multiple keys.

Items: max 100

permissionsstring[]#

Grants specific permissions directly to this key without requiring role membership.
Wildcard permissions like documents.* grant access to all sub-permissions including documents.read and documents.write.
Direct permissions supplement any permissions inherited from assigned roles.

Items: max 1000

expiresinteger#

Sets when this key automatically expires as a Unix timestamp in milliseconds.
Verification fails with code=EXPIRED immediately after this time passes.
Omitting this field creates a permanent key that never expires.

Avoid setting timestamps in the past as they immediately invalidate the key.
Keys expire based on server time, not client time, which prevents timezone-related issues.
Essential for trial periods, temporary access, and security compliance requiring key rotation.

Range: 0–4102444800000

creditsobject#

Credit configuration and remaining balance for this key.

Show child attributes

remainingintegerrequired#

Number of credits remaining (null for unlimited).

Range: 0–9223372036854776000

refillobject#

Configuration for automatic credit refill behavior.

Show child attributes

intervalenum<string>required#

How often credits are automatically refilled.

Options:dailymonthly

amountintegerrequired#

Number of credits to add during each refill cycle.

Range: 1–9223372036854776000

refillDayinteger#

Day of the month for monthly refills (1-31).
Only required when interval is 'monthly'.
For days beyond the month's length, refill occurs on the last day of the month.

Range: 1–31

ratelimitsobject[]#

Defines time-based rate limits that protect against abuse by controlling request frequency.
Unlike credits which track total usage, rate limits reset automatically after each window expires.
Multiple rate limits can control different operation types with separate thresholds and windows.
Essential for preventing API abuse while maintaining good performance for legitimate usage.

Items: max 50

Show child attributes

namestringrequired#

The name of this rate limit. This name is used to identify which limit to check during key verification.

Best practices for limit names:

Use descriptive, semantic names like 'api_requests', 'heavy_operations', or 'downloads'
Be consistent with naming conventions across your application
Create separate limits for different resource types or operation costs
Consider using namespaced names for better organization (e.g., 'files.downloads', 'compute.training')

You will reference this exact name when verifying keys to check against this specific limit.

Length: 3–128

limitintegerrequired#

The maximum number of operations allowed within the specified time window.

When this limit is reached, verification requests will fail with code=RATE_LIMITED until the window resets. The limit should reflect:

Your infrastructure capacity and scaling limitations
Fair usage expectations for your service
Different tier levels for various user types
The relative cost of the operations being limited

Higher values allow more frequent access but may impact service performance.

Range: >= 1

durationintegerrequired#

The duration for each ratelimit window in milliseconds.

This controls how long the rate limit counter accumulates before resetting. Common values include:

1000 (1 second): For strict per-second limits on high-frequency operations
60000 (1 minute): For moderate API usage control
3600000 (1 hour): For less frequent but costly operations
86400000 (24 hours): For daily quotas

Shorter windows provide more frequent resets but may allow large burst usage. Longer windows provide more consistent usage patterns but take longer to reset after limit exhaustion.

Range: >= 1000

autoApplybooleanrequired#

Whether this ratelimit should be automatically applied when verifying a key.

Default: false

enabledboolean#

Controls whether the key is active immediately upon creation.
When set to false, the key exists but all verification attempts fail with code=DISABLED.
Useful for pre-creating keys that will be activated later or for keys requiring manual approval.
Most keys should be created with enabled=true for immediate use.

Default: true

recoverableboolean#

Controls whether the plaintext key is stored in an encrypted vault for later retrieval.
When true, allows recovering the actual key value using keys.getKey with decrypt=true.
When false, the key value cannot be retrieved after creation for maximum security.
Only enable for development keys or when key recovery is absolutely necessary.

Default: false

Responses

application/json

Successfully created a new API key. The response includes both the keyId (for reference in your system) and the full key string. IMPORTANT: This is the only time the complete key is available - it cannot be retrieved later. You must securely provide this key to your end user immediately.

metaobjectrequired#

Metadata object included in every API response. This provides context about the request and is essential for debugging, audit trails, and support inquiries. The requestId is particularly important when troubleshooting issues with the Unkey support team.

Show child attributes

requestIdstringrequired#

A unique id for this request. Always include this ID when contacting support about a specific API request. This identifier allows Unkey's support team to trace the exact request through logs and diagnostic systems to provide faster assistance.

dataobjectrequired#

Show child attributes

keyIdstringrequired#

The unique identifier for this key in Unkey's system. This is NOT the actual API key, but a reference ID used for management operations like updating or deleting the key. Store this ID in your database to reference the key later. This ID is not sensitive and can be logged or displayed in dashboards.

keystringrequired#

The full generated API key that should be securely provided to your user.
SECURITY WARNING: This is the only time you'll receive the complete key - Unkey only stores a securely hashed version. Never log or store this value in your own systems; provide it directly to your end user via secure channels. After this API call completes, this value cannot be retrieved again (unless created with recoverable=true).

metaobjectrequired#

Show child attributes

requestIdstringrequired#

errorobjectrequired#

Base error structure following Problem Details for HTTP APIs (RFC 7807). This provides a standardized way to carry machine-readable details of errors in HTTP response content.

Show child attributes

detailstringrequired#

A human-readable explanation specific to this occurrence of the problem. This provides detailed information about what went wrong and potential remediation steps. The message is intended to be helpful for developers troubleshooting the issue.

statusintegerrequired#

HTTP status code that corresponds to this error. This will match the status code in the HTTP response. Common codes include 400 (Bad Request), 401 (Unauthorized), 403 (Forbidden), 404 (Not Found), 409 (Conflict), and 500 (Internal Server Error).

titlestringrequired#

A short, human-readable summary of the problem type. This remains constant from occurrence to occurrence of the same problem and should be used for programmatic handling.

typestringrequired#

A URI reference that identifies the problem type. This provides a stable identifier for the error that can be used for documentation lookups and programmatic error handling. When followed, this URI should provide human-readable documentation for the problem type.

errorsobject[]required#

List of individual validation errors that occurred in the request. Each error provides specific details about what failed validation, where the error occurred in the request, and suggestions for fixing it. This granular information helps developers quickly identify and resolve multiple issues in a single request without having to make repeated API calls.

Show child attributes

locationstringrequired#

JSON path indicating exactly where in the request the error occurred. This helps pinpoint the problematic field or parameter. Examples include:

'body.name' (field in request body)
'body.items[3].tags' (nested array element)
'path.apiId' (path parameter)
'query.limit' (query parameter)
Use this location to identify exactly which part of your request needs correction.

messagestringrequired#

Detailed error message explaining what validation rule was violated. This provides specific information about why the field or parameter was rejected, such as format errors, invalid values, or constraint violations.

fixstring#

A human-readable suggestion describing how to fix the error. This provides practical guidance on what changes would satisfy the validation requirements. Not all validation errors include fix suggestions, but when present, they offer specific remediation advice.

Add key roles Delete API keys