How obot-platform/obot works
Obot is an organizational platform for governing AI clients, model access, MCP tools, hosted workloads, and related activity. A typical user or client discovers approved MCP servers or skills through the CLI or UI, then connects through Obot’s gateways, where access and policy controls apply; activity can be audited across gateway and device sources. The README documents additional device-management, catalog, hosted-agent, identity, and export capabilities. Edges reflect those documented responsibilities; sampled source was limited and does not establish every subsystem’s wiring.
Main entry points: server.go API server · source.go Activity audit · root.go Obot CLI · handler.go MCP gateway
Components 21
Governed gateways
- MCP gateway
pkg/api/handlers/mcpgateway/handler.go - LLM gateway
pkg/gateway/server/llmproxy.go - MCP runtime
pkg/mcp/manager.go
Catalogs and workloads
- MCP and skills registries
pkg/api/handlers/registry/handler.go - Skills management
pkg/cli/skills.go - Hosted agents
pkg/agentbackend/backend.go
Identity and audit
- Model access policy
pkg/modelaccesspolicy/helper.go - Device governance
pkg/api/handlers/devicescans.go - Identity and access
pkg/api/authz/authz.go - Activity audit
pkg/auditlog/source.go - Audit export
pkg/auditlogexport/storage.go
Platform runtime
- API server
pkg/api/server/server.go - Resource controllers
pkg/controller/controller.go - Platform storage
pkg/storage/server.go
Other components
- Users and admins
- AI clients
- MCP servers
- Model providers
- Git providers
Connections 23
- Users and admins → Obot CLI: uses
- Users and admins → Web UI: uses
- AI clients → MCP gateway: connects to
- AI clients → LLM gateway: connects to
- Obot CLI → API server: requests services
- Web UI → API server: requests services
- API server → Identity and access: authorizes requests
- API server → Activity audit: records API activity
- MCP gateway → MCP runtime: manages sessions
- MCP gateway → MCP servers: proxies MCP traffic
- MCP gateway → Identity and access: enforces access
- LLM gateway → Model providers: proxies model requests
- LLM gateway → Model access policy: restricts models
- MCP and skills registries → Git providers: indexes catalogs
- Obot CLI → MCP and skills registries: discovers approved items
- Obot CLI → Skills management: manages skills
- Resource controllers → Hosted agents: reconciles workloads
- MCP runtime → Activity audit: records MCP activity
- LLM gateway → Activity audit: records model activity
- Device governance → Activity audit: contributes device activity
- Activity audit → Audit export: exports activity
- Resource controllers → Platform storage: reconciles resources
- API server → Platform storage: reads and writes