Agent report card
Can an agent use Anthropic API through API service lifecycle?
Grade D: Eligibility depends on information or judgment unavailable to an agent.
Anthropic / active
Five-stage assessment
Where agent autonomy holds and where it breaks
First break: Evaluate
Evaluate
BlockedEligibility depends on information or judgment unavailable to an agent.
Only part of the applicable terms is machine-readable.
An ARD declaration has not been established.
An agent can discover the product and funnel from the tested public entrypoints.
Pricing or the absence of a charge is explicit and machine-readable.
Evidence and actions
ARD declaration unknown
missing: unknownAn ARD declaration has not been established.
Decidable eligibility unavailable
supported: freshEligibility depends on information or judgment unavailable to an agent.
Eligibility under the Commercial Terms of Service depends on subjective or external factors, such as having legal authority to bind an entity, verifying identity/use upon request, and complying with usage policies.
Machine discovery
supported: freshAn agent can discover the product and funnel from the tested public entrypoints.
The product and funnel tools, SDKs, and API overview are fully discoverable via the public documentation entrypoint.
Readable pricing
supported: freshPricing or the absence of a charge is explicit and machine-readable.
Per-token pricing, feature multipliers, tool usage rates, and batch processing rates are explicitly disclosed and machine-readable on the pricing documentation page.
Structured terms limited
supported: freshOnly part of the applicable terms is machine-readable.
Commercial terms are available as stable web text with structured sections, but are provided as HTML/text rather than a dedicated structured data API.
What blocks agents
- Eligibility depends on information or judgment unavailable to an agent.
What would improve this stage
- Publish complete, explicit eligibility criteria and required inputs.
- Publish complete applicable terms in a stable machine-readable document.
Sign up
ReadyA stable signup entrypoint is available.
Whether phone verification is required has not been established.
The tested signup flow exposes machine-operable controls.
No CAPTCHA was encountered on the tested signup path.
A scoped, revocable machine identity flow is documented, including explicit human authorization boundaries.
Evidence and actions
CAPTCHA-free signup
supported: freshNo CAPTCHA was encountered on the tested signup path.
No CAPTCHA challenge was present on the observed login/signup path.
Machine-operable signup
supported: freshThe tested signup flow exposes machine-operable controls.
The sign in and sign up controls (such as 'Continue with Google', 'Continue with email', and 'Continue with SSO') are present in the DOM.
Phone-free signup unknown
supported: freshWhether phone verification is required has not been established.
Whether phone verification is required during the complete sign up flow is not established in the captured landing page evidence.
Stable signup entrypoint
supported: freshA stable signup entrypoint is available.
A stable signup and login entrypoint is available at the standard platform URL.
Supported identity flow
supported: freshA scoped, revocable machine identity flow is documented, including explicit human authorization boundaries.
Documentation supports both API keys from Console and scoped, short-lived OAuth access tokens via Workload Identity Federation.
Pay
BlockedThe required checkout cannot be operated by an agent.
A supported payment rail has not been established.
The required price or no-charge status is explicit before commitment.
The tested funnel can complete without sales intervention.
Evidence and actions
Explicit price
supported: freshThe required price or no-charge status is explicit before commitment.
Anthropic provides explicit pricing details for base input/output tokens, caching, and feature usage on their official pricing documentation.
Machine-operable checkout unavailable
supported: freshThe required checkout cannot be operated by an agent.
The login and authentication entry point for accessing console keys requires email, SSO, or Google interactive authentication mechanisms.
Sales-free purchase
supported: freshThe tested funnel can complete without sales intervention.
Self-service platform signup and API account creation can proceed via automated console login options without requiring sales contact.
Supported payment rail unknown
supported: freshA supported payment rail has not been established.
The console login endpoint does not expose documentation on payment rails directly within the captured login page state.
What blocks agents
- The required checkout cannot be operated by an agent.
What would improve this stage
- Expose documented, labelled checkout controls or an agent-compatible commerce interface.
Provision
ReadyUsable credentials or delegated access are delivered through a documented flow.
Automatic activation has not been established.
Machine-readable provisioning status has not been established.
A bounded provisioning delay has not been established.
Evidence and actions
Automatic activation unknown
supported: freshAutomatic activation has not been established.
Retained evidence from the console login page demonstrates authentication entry points (Google, email, SSO) but does not establish whether account onboarding automatically activates usable platform API access without manual intervention.
Bounded provisioning delay unknown
supported: freshA bounded provisioning delay has not been established.
Documentation provides standard API HTTP status codes and error shapes, but does not state a documented or bounded provisioning delay for API key or credential creation.
Credential delivery
supported: freshUsable credentials or delegated access are delivered through a documented flow.
API keys and delegated access (via Workload Identity Federation short-lived tokens) are delivered and generated through the web Console in Account Settings or POST /v1/oauth/token.
Machine-readable provisioning status unknown
supported: freshMachine-readable provisioning status has not been established.
Documentation describes synchronous HTTP API error response structures and status codes, but does not document a machine-readable provisioning progress, status, or reconciliation lifecycle interface.
Operate
LimitedRecovery is available only for some failure modes or requires human intervention.
An agent protocol interface has not been established.
A documented API or tool interface is available for operation.
Authentication is documented and usable within an agent authorization boundary.
Limits, structured errors, retry behavior, and idempotency semantics are documented where applicable.
Evidence and actions
Agent protocol interface unknown
missing: unknownAn agent protocol interface has not been established.
API or tool access
supported: freshA documented API or tool interface is available for operation.
Anthropic provides documented API and tool interfaces including client SDKs in seven languages and the ant CLI for interacting with the Claude API.
Delegated recovery limited
supported: freshRecovery is available only for some failure modes or requires human intervention.
SDKs and endpoints support token rotation and exponential backoff retry for transient errors, but full delegated recovery relies on human support intervention for persistent failures.
Documented limits and errors
supported: freshLimits, structured errors, retry behavior, and idempotency semantics are documented where applicable.
HTTP status codes, structured JSON error responses, request size limits, and SDK automatic retry behavior with exponential backoff are fully documented.
Usable authentication
supported: freshAuthentication is documented and usable within an agent authorization boundary.
Authentication via x-api-key or Bearer tokens obtained via Workload Identity Federation is fully documented and managed automatically by official SDKs.
What would improve this stage
- Provide delegated credential rotation, revocation, retry, and recovery paths without requiring unrestricted human intervention.
What works
- Sign up: A stable signup entrypoint is available.
- Provision: Usable credentials or delegated access are delivered through a documented flow.
What blocks agents
- Evaluate: Eligibility depends on information or judgment unavailable to an agent.
- Pay: The required checkout cannot be operated by an agent.
- Operate: Recovery is available only for some failure modes or requires human intervention.
Improve the report
Correct it, rerun it, or improve the funnel
Every request carries the exact profile revision, projection, declaration, and Catalog release.
Method, policy, and immutable identifiers
Complete, fresh evidence for the core service lifecycle first applies hard stage severity, then grades the share of applicable evidenced signals that are constrained. Optional blocker probes affect the grade when established and remain visible as unknown when public evidence cannot establish them.
Any incomplete, contradictory, unsupported, or non-fresh required evidence produces an unrated profile. A failed Evaluate, Sign up, or Pay stage derives D; a failed Provision or Operate stage derives F; otherwise constrained-signal ratio derives A+ through C.
- Profile ID
- arp_01kzf0m4x54qzpnr923g24x6x8
- Revision
- sha256:45f32e69a40a8a6c33793ac9543a0186800f2b6c0258951ba5299a0990305877
- Projection
- sha256:5e027ea3725855fff41828ee0b8c37a6f4a140818551f6a338b944c4028c9011
- Policy
- service-lifecycle-2026-08-08-calibrated: sha256:334e56852b5c38e178fcea839cfba680fb82dd96693842d933776baed21d63df
- Base Catalog release
- sha256:5082ec07047a2324648f569d234cc64342ce2707b2f3d77fde8fc06af517cee6