Plugins

Connectors that agents can actually be trusted with

A plugin is more than an API wrapper. It declares what it can do, how it authenticates and which actions need a human — so giving an agent access is a bounded decision.

Inside the product

Every connector, with its scopes on the card

This is the plugins directory in Enverif. Each card shows the publisher, the version, the number of agent actions and the exact capability scopes the connector asks for — before you create a connection.

Scoped connections. One plugin can hold several connections — different mailboxes, workspaces or accounts — each with its own credential.
Versioned packages. Plugins are published with a semantic version, so an agent's tool surface only changes when you upgrade it.
Named agent actions. Gmail exposes actions such as mail.account.read, mail.search, mail.thread.read and mail.message.read — an agent cannot invent a call the plugin never declared.

Plugin documentation Extend with MCP

Enverif plugins directory listing Apify, Apollo, webhook automation, Buffer, Calendly, Gmail, Google Analytics, Google Maps, Search Console, Google Sheets, Outlook and Slack connectors — The plugins directory inside Enverif. Every connector declares its agent actions and capability scopes before an agent can use it.

Capability scopes

Six scopes, printed on every plugin

Scopes are the contract between a connector and an agent. They are declared in the manifest, shown in the UI and enforced at call time.

read

Fetch data only. Inbox search, sheet reads, analytics queries and lead lookups all sit here.

internal_write

Change data inside your Enverif workspace — leads, notes, campaign records — without touching the outside world.

external_write

Anything that leaves your workspace: sending mail, posting to Slack, publishing a scheduled post. Gated by approval unless you explicitly enable autonomous sending.

network

Outbound HTTP to a declared endpoint, used by webhook and scraping plugins.

secrets

Access to a stored credential for the connection the plugin was authorised against — never another connection's.

destructive

Deletes and overwrites. Always visible on the plugin card before you connect it.

Anatomy

What ships inside a plugin

Manifest

Declares the plugin's name, publisher, category and the capabilities it exposes to agents.

Authentication

OAuth or API key, authorised per connection with the narrowest scopes the provider supports.

Capabilities

Explicit, named operations. An agent can only call what the plugin declares.

Approval policy

Write and send operations are gated by default, so a human sees the payload first.

Run logging

Every call is recorded with its inputs, outputs and the agent that made it.

Built in

13 plugins ship with Enverif

Catalog 1.0.0, generated 2026-08-08 from ShubhamTuts/enverif.

Google Search Console

Full integration catalog Contribute a plugin

AI models

Plugins handle the tools. Provider adapters handle the thinking.

Model access is its own connection type. Add a provider key once, then choose the model per agent — the plugin surface stays exactly the same.

OpenAI
Provider adapter in Enverif source. Bring your own API key or use managed credits.
Provider docs
Anthropic
Provider adapter in Enverif source. Bring your own API key or use managed credits.
Provider docs
Google Gemini
Provider adapter in Enverif source. Bring your own API key or use managed credits.
Provider docs
DeepSeek
Provider adapter in Enverif source. Bring your own API key or use managed credits.
Provider docs

Custom model IDs are supported on every adapter, and a connection can be tested before an agent is pointed at it.

Connect the tools your team already pays for.

Every built-in plugin ships with scoped credentials, declared actions and approval gates on anything that leaves your workspace.

Join the waitlist Get a demo