Skip to content
CantoraAPI documentation
POST
/v1/organizations/{organizationId}/projects/{projectId}/agent-configuration-previews

Preview one complete Agent Configuration without changing Product State

Return the canonical actions, manifest digest, warnings, and state preconditions for one complete desired Test Agent Configuration without changing Product State.

Authentication

Send an API key as a Bearer token in the Authorization header.

Parameters

NameLocationRequiredDescription
organizationIdpathYes
The Organization to address.
string
All of
  • maximum length 64
A Cantora Organization identifier, prefixed with `org_`.
  • pattern ^org_[\s\S]+$
projectIdpathYes
The Project to address.
string
All of
  • maximum length 64
A Cantora Project identifier, prefixed with `proj_`.
  • pattern ^proj_[\s\S]+$

Request body

Required.

application/json
object
  • unknown properties allowed false
desiredrequired
DesiredAgentConfigurationThe complete desired Test Agent Configuration to evaluate without writing.

Responses

200AgentConfigurationPreview
application/json
AgentConfigurationPreviewJsonEncoding
400The request path, headers, query, or JSON body did not satisfy the published schema
401Unauthorized
application/json
UnauthorizedJsonEncoding
403Forbidden
application/json
ForbiddenJsonEncoding
404NotFound
application/json
NotFoundJsonEncoding
409Conflict
application/json
ConflictJsonEncoding

Reusable schemas

AgentConfigurationPreviewJsonEncoding

object
  • unknown properties allowed false
manifestDigestrequired
stringThe lowercase SHA-256 digest of the normalized manifest.
actionsrequired
arrayThe ordered configuration actions required to converge Product State.
Items
string
  • allowed values "createAgentDefinition", "updateAgentDefinition", "createTestAgent", "publishAgentVersion", "reuseAgentVersion", "activateTestRelease", "reuseTestRelease"
preconditionsrequired
AgentConfigurationPreconditionsThe exact mutable Product State observed during preview.
warningsrequired
arrayNon-blocking conditions the caller should review before applying.
Items
AgentConfigurationWarningJsonEncoding

UnauthorizedJsonEncoding

object
  • unknown properties allowed false
_tagrequired
stringThe stable machine-readable error type.
  • allowed values "Unauthorized"
reasonrequired
stringA safe explanation of why the credential was rejected.

ForbiddenJsonEncoding

object
  • unknown properties allowed false
_tagrequired
stringThe stable machine-readable error type.
  • allowed values "Forbidden"
permissionrequired
stringThe permission required by the refused operation.

NotFoundJsonEncoding

object
  • unknown properties allowed false
_tagrequired
stringThe stable machine-readable error type.
  • allowed values "NotFound"
resourcerequired
stringThe resource type relevant to the error.
idrequired
stringThe identifier supplied for the resource that was not found.

ConflictJsonEncoding

object
  • unknown properties allowed false
_tagrequired
stringThe stable machine-readable error type.
  • allowed values "Conflict"
resourcerequired
stringThe resource type relevant to the error.
reasonrequired
stringA safe explanation of the state conflict.

DesiredAgentConfiguration

object
  • unknown properties allowed false
intentrequired
stringWhether the configuration creates a new binding or updates an identified Agent Definition.
  • allowed values "create", "update"
agentDefinitionIdrequired
The Agent Definition identifier.
Any of
string
All of
  • maximum length 64
A Cantora Agent Definition identifier, prefixed with `agentdef_`.
  • pattern ^agentdef_[\s\S]+$
null
keyrequired
stringThe stable, human-authored Agent Definition key.
displayNamerequired
stringThe human-readable resource name.
manifestrequired
AgentVersionManifestThe complete immutable Agent behavior and execution-budget declaration.
testReleaserequired
objectThe desired Test Environment and activation reason.
  • unknown properties allowed false
environmentIdrequired
stringThe Environment identifier.
All of
  • maximum length 64
A Cantora Environment identifier, prefixed with `env_`.
  • pattern ^env_[\s\S]+$
reasonrequired
stringThe human-readable reason for activating the Test Agent Version.

AgentConfigurationPreconditions

object
  • unknown properties allowed false
agentDefinitionIdrequired
The Agent Definition identifier.
Any of
string
All of
  • maximum length 64
A Cantora Agent Definition identifier, prefixed with `agentdef_`.
  • pattern ^agentdef_[\s\S]+$
null
definitionKeyrequired
The stable Agent Definition key.
Any of
string
null
definitionDisplayNamerequired
The Agent Definition display name observed during preview, or null when creating it.
Any of
string
null
definitionStatusrequired
The Agent Definition lifecycle status observed during preview, or null when creating it.
Any of
string
  • allowed values "active", "disabled"
null
configurationAuthorityrequired
The one surface allowed to change this Agent Definition and publish its Versions.
Any of
string
  • allowed values "ApiManaged", "DashboardManaged"
null
testAgentPrincipalIdrequired
The Test Agent Principal identifier, or null before it is created.
Any of
string
All of
  • maximum length 64
A Cantora Principal identifier, prefixed with `principal_`.
  • pattern ^principal_[\s\S]+$
null
testReleaseGenerationrequired
integerThe current Test Agent Release generation, starting at zero before activation.
All of
  • minimum 0
currentManifestDigestrequired
The current Test Agent Version manifest digest, or null before the first activation.
Any of
string
null

AgentConfigurationWarningJsonEncoding

object
  • unknown properties allowed false
coderequired
stringThe stable machine-readable condition code.
messagerequired
stringThe safe human-readable condition detail.

AgentVersionManifest

object
  • unknown properties allowed false
schemaVersionrequired
stringThe Agent Version manifest schema identifier.
  • allowed values "config.cantora.ai/agent-version/v1alpha2"
instructionsrequired
stringThe complete natural-language instructions for the Agent Version.
All of
  • maximum length 262144
modelrequired
AgentVersionModelThe exact provider model and behavior settings pinned by the Agent Version.
toolsrequired
AgentVersionDependenciesThe tool keys the Agent Version may call.
requiredBindingsrequired
AgentVersionDependenciesThe configuration bindings the Agent Version requires at runtime.
budgetsrequired
AgentVersionBudgetsThe finite execution limits enforced for the Agent Version.

AgentVersionModel

Any of
object
  • unknown properties allowed false
providerrequired
stringThe model or Surface provider key.
  • allowed values "googleVertex"
modelrequired
stringThe exact provider model and behavior settings pinned by the Agent Version.
All of
  • minimum length 1
  • maximum length 256
settingsrequired
GoogleVertexModelSettingsThe provider-native settings that change model behavior.
object
  • unknown properties allowed false
providerrequired
stringThe model or Surface provider key.
  • allowed values "openai"
modelrequired
stringThe exact provider model and behavior settings pinned by the Agent Version.
All of
  • minimum length 1
  • maximum length 256
settingsrequired
OpenAiModelSettingsThe provider-native settings that change model behavior.

AgentVersionDependencies

array
Items
string
All of
  • minimum length 1
  • maximum length 100
  • pattern ^[a-z][a-z0-9_]*$
All of
  • at most items 100
  • unique items true

AgentVersionBudgets

object
  • unknown properties allowed false
maxModelStepsrequired
integerThe maximum number of model calls in one Run.
All of
  • maximum 100
  • minimum 1
maxToolCallsrequired
integerThe maximum number of tool calls in one Run.
All of
  • maximum 100
  • minimum 0
totalWallTimeMsrequired
integerThe maximum total Run duration in milliseconds.
All of
  • maximum 300000
  • minimum 1000
modelStepTimeMsrequired
integerThe maximum duration of one model step in milliseconds.
All of
  • maximum 300000
  • minimum 1000
streamStallTimeMsrequired
integerThe maximum duration without stream progress in milliseconds.
All of
  • maximum 300000
  • minimum 1000
maxInputTokensrequired
integerThe maximum number of model input tokens.
All of
  • maximum 10000000
  • minimum 1
maxOutputTokensrequired
integerThe maximum number of model output tokens.
All of
  • maximum 1000000
  • minimum 1
maxReasoningTokensrequired
integerThe maximum number of model reasoning tokens.
All of
  • maximum 1000000
  • minimum 0
maxTotalTokensrequired
integerThe maximum total model tokens in one Run.
All of
  • maximum 10000000
  • minimum 1
maxToolResultBytesrequired
integerThe maximum aggregate tool-result size in bytes.
All of
  • maximum 10000000
  • minimum 1
maxEstimatedCostMicrosrequired
integerThe maximum estimated model cost in millionths of the billing currency.
All of
  • maximum 1000000000
  • minimum 1

GoogleVertexModelSettings

object
  • unknown properties allowed false
thinkingLevelrequired
stringThe Google Vertex-native thinking level.
  • allowed values "minimal", "low", "medium", "high"

OpenAiModelSettings

object
  • unknown properties allowed false
reasoningEffortrequired
stringThe OpenAI-native reasoning effort.
  • allowed values "none", "low", "medium", "high", "xhigh", "max"
reasoningModerequired
stringThe OpenAI-native reasoning mode.
  • allowed values "standard", "pro"