AgentValet vs 1Password

1Password is where a lot of teams already keep their secrets, and it now positions several products for AI agents. This page compares Service Accounts, the product 1Password’s developer docs describe for authenticating automated workflows and production agents, with AgentValet. They answer different questions about the same key.

Updated

The short answer

Pick AgentValet if your AI agents act on platforms such as Slack, GitHub or Stripe and should never receive the credential, with a person able to approve risky actions first.

Pick 1Password Service Accounts if your team already keeps its secrets in 1Password and you need scripts, pipelines or agents to read those secrets with a scoped token that isn’t tied to a person.

Two approaches, in plain words

What 1Password Service Accounts are

A 1Password service account is a token, not tied to any person, that authenticates the 1Password CLI and SDKs (Go, JavaScript and Python). An administrator chooses which vaults and Environments it can reach and whether it can read, write or share items, and those choices can’t be changed after creation. Tokens can be given an expiry, and usage reports show which items a service account accessed.

1Password positions other pieces for AI agents too. Credential Broker, in public preview on 1Password Business, verifies a workload by OIDC token (GitHub Actions or a custom provider) and makes secrets available for the duration of the workflow. Agentic Autofill, built with Browserbase, lets agents sign in to websites. Local agent brokering and agent identity attestation are listed as coming soon.

What AgentValet is

AgentValet is a credential broker for AI agents. The agent never reads the secret: it asks the broker to make the call, the broker checks the agent’s grant for that platform and action, and attaches the credential in memory.

A risky action can wait for a person, and each decision is logged with the policy, version, result and reason. AgentValet doesn’t store passwords for your team or fill logins in a browser.

The question that separates them

Once the agent has read the secret, what governs what it does with it?

A service account governs which secrets can be read. After that, the secret is a value in the agent’s process and the platform sees an ordinary API key. 1Password’s own tutorial on using its SDKs with AI agents says it isn’t the recommended integration approach, and that exposing raw credentials directly to an AI model carries significant risks.

AgentValet governs the call instead of the read. Each request is matched against the agent’s grant for that platform and action before the credential is decrypted, and the key never leaves the broker.

A service account governs the read
  1. Agent with a service account token
  2. 1Password checks vault access
  3. Secret value returned to the agent’s process (the key sits with the agent)
  4. Agent calls the platform
  5. Real platform

Vault access is scoped and reported on. What the agent then does with the key is outside 1Password’s view.

AgentValet enforces at the platform call
  1. Your agent
  2. Broker checks identity, grant and approval
  3. Credential attached in memory
  4. SaaS API or MCP server

The agent holds no platform key, and a risky call can wait for a person before it runs.

The standard comparison

One line per cell. Open the detail underneath for the longer version.

AgentValet compared with 1Password Service Accounts
AgentValet1Password Service Accounts
Per-agent identityEach agent gets its own RS256 signing keypair and signs every request.A service account per workload, up to 100 per account, each with its own token.
Credential exposure to the agentThe agent never holds or sees the credential.The process using the token reads the secret value itself.
Token lifetimeEach request carries a short-lived token signed by the agent; upstream credentials stay in the broker.Service account tokens can be given an expiry when created, and can be rotated or revoked.
Scoping granularityPer agent, per platform and per action, deny by default.Per vault and Environment, with read, write and share permissions fixed at creation.
Human approvalRisky actions pause at the broker until a person approves them.Not described for service accounts; access is set by an administrator when the account is created.
Audit evidenceEvery decision is logged with the policy, version, result and reason, and exports as CSV or JSON.Usage reports show which items a service account accessed.
MCP supportA native MCP server, and it brokers calls to upstream MCP servers as an OAuth client.An MCP server for 1Password Environments lets agents create and manage Environments through authorisation prompts.
Self-host optionNo. It is a managed service hosted in Australia.No for 1Password itself; Connect servers can run in your infrastructure and cache your data there.
Open sourceThe parts that run on your machines (MCP server, CLI, SDKs and adapters) are MIT licensed. The broker is not.The SDKs for Go, JavaScript and Python are MIT licensed; the 1Password service isn’t open source.
Setup timeOne command, npx @agentvalet/register, registers an agent.Create a service account on 1Password.com or with the CLI, then export its token as OP_SERVICE_ACCOUNT_TOKEN.
Starting priceFree plan, with paid plans from US$99 a month.1Password Business is US$8.99 per user per month, billed yearly, with a 14-day free trial; Enterprise is quoted.
The detail behind 7 of these rows
Per-agent identity

AgentValet: Every agent publishes an identity document and key set, so each call is attributable to one agent. Child agents can be issued their own identities with scopes cut down to a subset of the parent’s.

1Password Service Accounts: Service accounts aren’t tied to an individual, so access is attributed to the service account rather than a person. Credential Broker, in preview, verifies a workload’s identity from OIDC token claims instead.

Credential exposure to the agent

AgentValet: Each credential is envelope-encrypted and decrypted in memory only for the call it serves. It is never logged or returned to the agent.

1Password Service Accounts: 1Password’s AI agent tutorial passes resolved credentials to the agent and warns that exposing raw credentials to a model carries significant risks, advising short-lived, tightly scoped tokens where possible.

Token lifetime

1Password Service Accounts: A token belonging to a deleted service account can’t authenticate. With Credential Broker, secrets stay accessible only for the duration of the workflow.

Scoping granularity

AgentValet: The method and endpoint of each call are matched against the scope granted to that agent before any credential is decrypted.

1Password Service Accounts: Built-in Personal, Private and Employee vaults and the default Shared vault can’t be granted, and Environments are read-only to service accounts.

Human approval

AgentValet: Approve with a passkey, an email link or a mobile push. The approval is recorded with the action it released.

Audit evidence

AgentValet: The log is append-only and policy-stamped, and each decision carries a signed receipt.

1Password Service Accounts: Credential Broker records every access request in the audit log.

Starting price

1Password Service Accounts: Service account rate limits depend on the plan: on Business, 10,000 reads and 1,000 writes an hour per token and 50,000 requests a day per account; on Teams, 5,000 requests a day.

Where 1Password genuinely wins

If your team already lives in 1Password, service accounts are the shortest path to getting secrets out of .env files and into scripts, CI and agents. The vaults your people already use become the source for your automation, with the same admin console, vault permissions and usage reports, and the CLI’s op run, op read and op inject commands cover most workflows.

1Password also covers ground AgentValet doesn’t: password management for people, browser sign-in for agents through Agentic Autofill, device trust, and Credential Broker for OIDC-verified workflows such as GitHub Actions.

Where AgentValet wins

AgentValet governs what the agent does with the credential, a layer a vault doesn’t reach. Each agent has its own signing identity, each call is matched against a per-platform, per-action grant before anything is decrypted, and a risky action can wait for a person to approve it with a passkey, an email link or a push notification.

Because the agent never holds the key, revoking the agent stops it on its next call, and each decision is logged with its policy and a signed receipt.

They fit together. Keep people’s passwords and your pipeline secrets in 1Password, and let AgentValet hold the platform credentials your AI agents act with.

The first step, side by side

Each command is the documented starting point for that product. 1Password’s comes from 1Password’s guide to getting started with service accounts.

AgentValet
npx @agentvalet/register
1Password (from its service accounts guide)
op service-account create <serviceAccountName> --expires-in <duration> --vault <vault-name:<permission>,<permission>

Which one fits

Choose 1Password if

  • Your team already uses 1Password and wants one place for human and machine secrets.
  • Your automation needs to read secrets into scripts, CI or containers.
  • You want agents to sign in to websites through a browser.

Choose AgentValet if

  • Your agents call Slack, GitHub, Stripe or MCP servers and shouldn’t hold the key.
  • You want a person to approve individual risky actions before they run.
  • You need a decision log that shows which agent did what, under which policy.

Run both if

  • Your people and pipelines use 1Password, and your AI agents act on SaaS platforms.

Govern one agent first.

Register an agent on the free plan with npx @agentvalet/register, grant it one scope, then try a call outside it. You’ll see the refusal, and the audit row it leaves behind.

npx @agentvalet/register