Openfort vs Agent Vault by Self

Openfort

Backend wallets sign inside a GCP Confidential Space enclave, secret keys carry 26 scopes, and rate limits, errors and prices are published. On EVM, a backend send reaches the policy engine only as a hash, so address and value rules bind only when the caller runs the pre-flight check first.

Agent Vault by Self

Scoped, revocable keys, OAuth on the hosted MCP server, spending caps, approvals and a documented threat model are designed to keep real card numbers and passwords out of the agent's context. The home page and launch post say cards are revealed only in an attested enclave, while the docs say that path is off on the hosted service.

Openfort

1. Call policies.evaluate with operation signEvmTransaction before every EVM backend send. The send itself is checked only as signEvmHash
2. Keep the signing policy and the gas sponsorship policy separate. Linking a signing policy to a fee sponsorship stops it working as a guardrail
3. Pass a fee sponsorship on EVM sends. Without one the transaction stays pending with no error
4. Give an agent a secret key without accounts:export, and limit the CLI MCP server to the tools it needs
5. On a 5xx after a write, read the resource before retrying. On a 429, wait the full Retry-After

Agent Vault by Self

1. Pass the session handle from each hosted MCP result on every later call. Without it a page tool is refused or a new signed-out browser opens
2. On a pending answer, call get_pending_result with its wait id in the same turn. Repeating the original call starts a second one
3. Tell the user when waiting_on_owner is true. A first fill on a new site waits up to 15 minutes for the holder's approval
4. Before retrying a fill after an unknown outcome, check the merchant's order state or get_run_report. Some failures keep a recorded spend
5. Ask the holder to add an agent card to every card the agent may use. A card marked No limits accepts any amount on authorised sites