CAIPE Features

Everything you need to run AI-powered platform engineering at enterprise scale — from multi-agent orchestration to RAG knowledge bases, skills, and production-grade deployment.

Get Started →View Roadmap Submit a Feature Request ↗

See it in action

Explore the two workflows that make CAIPE useful from the first day.

Agent Builder

Configure agents, tools, knowledge, skills, and runtime guardrails from one guided workspace.

Read the docs →Watch on Vidcast ↗

Knowledge Bases

Ingest sources, search governed content, and organize reusable collections for agents.

Read the docs →Watch on Vidcast ↗

🛠️

Agent Builder

Six-step builder for identity, instructions, tools, knowledge, skills, and advanced controls
Configured models, MCP tools, built-in tools, data sources, and collections
Subagents, workflows, middleware, and human approval rules
Team ownership and global sharing with RBAC enforcement
Test in chat, clone, export, or bootstrap agents through Helm

🔁

Workflows

Ordered agent steps with per-step prompts and optional capability overrides
Jinja-compatible context from previous steps and run input
Abort, skip, or retry failed steps; pause for human input or approval
Persistent run timelines, tool events, statuses, and artifacts
Start runs from the UI, an agent with workflow access, or the REST API

🔌

BYO Agents & MCP Servers

Deploy code-defined agents and connect external MCP servers
appConfig.mcp_servers — plug in external MCP servers at chart install time
Per-tool MCP servers for each integration
Supported MCP transports: stdio, SSE, and Streamable HTTP
Credential sources: secret refs, caller tokens, and provider connections
LiteLLM proxy support — any LLM provider through a single endpoint

🧩

External Apps

Publish independently deployed web apps in the CAIPE Apps hub
Reuse the signed-in CAIPE session for app-scoped access
Protect routes with explicit roles and token scopes
Keep app data, UI, and domain authorization independent
Same-origin hosting with a short-lived identity token per request

🔑

Credentials & Secrets

Connected Apps — link Atlassian, GitHub, GitLab, PagerDuty, and Webex for user-scoped MCP access
Connection health, reconnect, and permission review from one Credentials page
Saved Secrets — store bearer tokens and other secret types without showing values again
Rotate, share, and revoke secrets with team-scoped RBAC
Server-side credential retrieval for agents and MCP servers — raw secrets never return to the browser

🤖

Multi-Agent Orchestration

Multi-agent and deep agent interactions with access to multiple tools and sub-agents based on customizable system prompts
10+ first-party curated sub-agents and MCP servers
Create agents with Agent Builder
Ability to customize system prompts
[Middleware] Custom Skills Integration

🎨

Rich Web UI

Rich/Contextual Home Page
Rich Chat Interface with live agent/tool status via streaming
Share chat with teams · Archive/Delete chats
Agent Builder
Skills Gateway — AI Assist, API access, security scanner, GitHub crawling

🧰

Skills

Browse and reuse focused procedures stored as SKILL.md files
Import, revise, and attach skills to agents
Scan skills for prompt injection and unsafe tool behavior
Use scan gates to warn or block flagged skills
Install catalog skills into supported coding agents through the Gateway

⚙️

Settings & Admin

Set personal defaults for chat, agents, appearance, and notifications
Review platform health and component status
Manage users, teams, resources, integrations, and credentials
Review audit, metrics, security policy, and migrations
Keep administrative controls separate from personal preferences

🧠

Knowledge Bases

Self-service file, web, and collaboration data sources
Deployment-managed infrastructure ingestors
Hybrid semantic and keyword retrieval with optional Graph RAG
Ownership and search access enforced through OIDC and OpenFGA
Reusable collections and MCP tools for search, fetch, discovery, and graph exploration

💾