Extend your AI systems with external capabilities. Define tool schemas, configure authentication, connect to any API or service, and invoke them directly from your augmentation steps.
Tools are first-class objects in MechaMental that can be invoked from any pipeline step. Define your tool schemas, configure authentication credentials, and call them from any augmentation — making it easy to integrate external capabilities into your AI systems.
Define input and output schemas for each tool so the platform validates calls at runtime and LLMs know how to invoke them correctly.
Configure API keys, OAuth tokens, or custom auth headers per tool. Credentials are stored securely in the workspace vault.
Once defined, tools are available to every pipeline in the workspace. Reference them in augmentation steps to invoke external systems inline.
MechaMental supports a variety of tool types to cover different integration patterns and use cases.
Call any REST API endpoint. Configure method, headers, body templates, and authentication for outbound HTTP requests.
Connect to Model Context Protocol servers for standardized tool access. Discover and invoke tools from any MCP-compliant provider.
Query databases directly from pipeline steps. Execute read operations against configured data sources with parameterized queries.
Run custom logic as tool invocations. Execute sandboxed code for data transformation, validation, or any compute task.
Integrate with SaaS platforms, cloud services, and external systems through pre-built connectors or custom API configurations.
Actions are automated workflows that run outside of user-initiated pipeline calls. MechaMental supports two types of actions, both with full execution history tracking.
Event-driven execution against endpoints. When a specific event occurs — such as a deployment completing, a document being ingested, or an external webhook arriving — trigger actions fire automatically and call the configured endpoint.
Cron-based recurring jobs that run on a defined schedule. Use scheduled actions for periodic data syncs, report generation, health checks, or any task that needs to run at regular intervals.
Both action types track a full execution history. Each run records its status so you can monitor, debug, and audit every automated workflow.
Bridges are WireGuard-based secure tunnels that let MechaMental call tools running in private networks or on-premise systems. The data plane runs in your infrastructure while the control plane stays managed — giving you connectivity without exposing internal services to the public internet.
MechaMental orchestrates the connection, manages WireGuard keys, and routes tool calls through the encrypted tunnel to your private infrastructure.
A lightweight bridge agent runs inside your network, establishing an outbound WireGuard tunnel. No inbound ports needed — your tools stay private.
Tools integrate seamlessly into augmentation pipelines. Each tool call is a step in the pipeline, and tools can be chained together for complex automation workflows.
Tools can be chained in augmentation pipelines for complex automation workflows. Each step can invoke one or more tools, pass results to the next step, and branch logic based on tool responses.