Design reusable agents, callable tools, visual missions, knowledge skills, and customer-facing channels in one Rails-native platform with built-in chat, observability, and multi-tenant operations.
Main features
The entry point to build, inspect, and operate AI systems.
Starting from the dashboard is the fastest path into agents, tools, missions, runs, and the rest of the admin features.
Configure reusable AI agents with instructions, model settings, tools, skills, subagents, and capabilities. Agents power playground chats, internal automation, and published channels.
Turn SQL, MCP, RAG, and mission flows into callable runtime capabilities. Tools keep data access and external actions explicit, assignable, and reviewable from the admin surface.
Design multi-step workflows with visual nodes, control flow, HTTP calls, tools, state, and outputs. Missions turn isolated prompting into repeatable orchestration with runtime visibility.
Full platform
The surrounding platform that turns those core surfaces into something teams can ship, inspect, and operate.
Reusable runtime units with instructions, tools, skills, subagents, and capability plugins.
Callable capabilities backed by queries, retrieval, integrations, and mission execution.
Progressive knowledge libraries that agents can discover and activate at runtime.
Visual orchestration for control flow, tool usage, state, and structured outputs.
Publish agents or missions through branded chat, API, and messaging entry points with scoped targets.
Retrieval pipelines for ingest, chunking, embedding, and grounded search over your content.
A fast operator surface for trying agents, inspecting responses, and watching live tool usage.
A built-in admin assistant that can inspect resources and delegate to designer subagents inside the app.
Runtime trace views for browsing chats, messages, tool calls, and execution history after the fact.
Execution visibility for mission runs, including status, history, and runtime progression.
Repeatable evaluation flows for validating agent and mission behavior before rollout.
Trusted connections to providers, databases, authentication systems, and external services.
Tenant-local workspaces for grouping agents, tools, missions, channels, and RAG flows by intent.
Extensible architecture for loading tool, connector, capability, and RAG modules into the app.
Top-level isolation for customer spaces, core resources, tenant-specific administration, and local logins.
Get started
Undercover Agents is open source, Rails-native, and designed for teams that need one place to build and operate AI systems.
