Docs / Core Concepts / Tools

Tools

Tools are the external capabilities Ploton connects to — databases, APIs, OAuth services, shell commands, and anything your agents need to act in the real world.

What are tools?

A tool is anything Ploton can call on behalf of your agent. Your database, your CRM, a payment processor, a shell command that runs ImageMagick, an internal API behind your VPN — all tools.

Tools are just capabilities. They expose functionality, Ploton connects to them, handles auth where needed, and calls them during task execution.

Types of tools

OAuth services

Third-party platforms that require user authorization — CRMs, payment processors, communication platforms, cloud storage.

Ploton manages the OAuth lifecycle: consent flows, token storage, refresh, scope management. Your agent never touches a raw token.

APIs

Any HTTP endpoint your agent needs to hit — REST, GraphQL, internal microservices. Auth can be API keys, bearer tokens, or nothing.

Databases

PostgreSQL, MySQL, MongoDB. Ploton can query, insert, update, and export data. You store connection credentials in Ploton’s config.

Shell commands

System-level tools like ImageMagick, FFmpeg, or custom scripts. Ploton runs these in a sandboxed environment as part of a task workflow.

Custom integrations

Anything that doesn’t fit the categories above. If it has an interface — HTTP, CLI, SDK — Ploton can wrap it as a tool.

How tools get used

Tools don’t run on their own. They get selected and invoked as part of a task’s workflow.

flowchart LR
    A["Agent prompt"] --> B["Analyze intent"]
    B --> C["Select tools"]
    C --> D["Authenticate"]
    D --> E["Execute"]
    E --> F["Return result"]

    style A fill:#1a1630,stroke:#FACC15,color:#e8e0f0
    style B fill:#1a1630,stroke:#FACC15,color:#e8e0f0
    style C fill:#1a1630,stroke:#3B82F6,color:#e8e0f0
    style D fill:#1a1630,stroke:#FACC15,color:#e8e0f0
    style E fill:#1a1630,stroke:#FACC15,color:#e8e0f0
    style F fill:#1a1630,stroke:#50FA7B,color:#e8e0f0

Say your agent sends a prompt like:

Pull all contacts from user_123's CRM account created this month

Ploton will:

  1. Analyze the prompt
  2. Identify the CRM as the relevant tool
  3. Build a workflow: authenticate, query contacts, filter by date, format response
  4. Execute each step, calling the CRM tool with the right parameters

A single task can use multiple tools. “Pull the user’s invoices from the payment service and upload a summary to their cloud storage” hits two tools in one workflow.

Built-in tools

Ploton ships with connectors for popular services:

CategoryTools
CRMHubSpot, Salesforce, Pipedrive
CommunicationSendGrid, Slack, Discord
StorageGoogle Drive, Dropbox, AWS S3
DataPostgreSQL, MySQL, MongoDB
PaymentsStripe, PayPal

We add more over time. For services not yet supported, you can register your own — see the Registering Custom Tools guide.

Tools vs. trained tasks

Tools and Trained Tasks are related but different:

  • A tool is a capability — a CRM API, a payment API, a database
  • A trained task is a task that benefits from execution memory — what Ploton learned from running past tasks against that tool

Every tool starts as a connector. As tasks run through it, Ploton accumulates execution memory — what worked, what failed, how to recover. That memory is what makes future tasks “trained.” Essentially, trained tasks run on tools that have seen enough real traffic to have useful history.

Tool:    "I can connect to this CRM"
Trained: "I can connect to this CRM, and I know that rate limits
          hit at 100 req/10sec, that the schema changed in the
          latest API version, and that token refresh takes ~200ms"

Next steps