litellm - 💡(How to fix) Fix Feature: Agent Identity Headers for LiteLLM Proxy — WTRMRK Integration [1 participants]

TL;DR

Implementing support for the X-Agent-UID header in LiteLLM Proxy, optionally verified against WTRMRK, can enable verifiable agent identity for downstream systems.

Guidance

• To integrate WTRMRK verification, ensure the litellm_config.yaml file is correctly configured with the agent_identity settings, including the provider, verify_on_request, and wtrmrk_api fields.
• Verify that each agent has a registered UID and cryptographic keypair on the Base network, as required by the WTRMRK protocol.
• Test the integration by sending LLM calls with the X-Agent-UID header and verifying that downstream systems can correctly authenticate and authorize the requests.
• Consider implementing per-agent cost tracking, verified agent audit trails, and trust-gated model access using the verified agent identities.

Example

# litellm_config.yaml
agent_identity:
  provider: wtrmrk
  verify_on_request: true
  wtrmrk_api: "https://wtrmrk.io/api/attestations"

Notes

This solution assumes that the WTRMRK protocol and API are correctly implemented and functioning as expected. Additionally, the integration may require modifications to the LiteLLM Proxy code to support the X-Agent-UID header and WTRMRK verification.

Recommendation

Apply workaround by implementing the proposed X-Agent-UID header support in LiteLLM Proxy, as it enables verifiable agent identity and provides a foundation for per-agent cost tracking, verified agent audit trails, and trust-gated model access.