Personal access token
A personal access token (PAT) is a credential a user generates for themselves to authenticate API calls without providing a password. PATs are scope-limited, revocable, and the standard pattern for programmatic access to GitHub, Notion, Linear, and most modern SaaS APIs.
Why it matters
Personal access tokens emerged from a real security problem: users who wanted programmatic access to an API used to have to type their password into a script. If the script leaked, so did the password. PATs solve this by giving the user a separate credential they generate per-purpose, scope as narrowly as needed, and revoke without changing their password.
For AI agents, PATs are how users wire personal tools into the agent without surrendering broader account access. The user generates a PAT in GitHub (or wherever) with the specific scopes the agent needs. The agent uses the PAT for tool calls; if the agent is compromised, the user revokes that single PAT and the agent loses access without affecting the user's password or other tokens.
The pattern works because most modern SaaS issues PATs with explicit scope-listing. Users see exactly what permissions the token grants before they create it. Revocation is one click. Audit logs show which actions used which token.
How Ajeris handles this
Ajeris's custom MCP integration supports user-generated PATs for any HTTPS API. The user generates a PAT in the source SaaS, pastes it into the agent's wizard, and the agent treats it as a bearer token for that specific tool.
PATs are stored encrypted at rest, per-user isolated, and never exposed to other users or to the model in plaintext. When a user revokes a PAT in the source SaaS, the next tool call returns 401; the agent surfaces this as a tool failure for the user to re-authenticate.
