Nimbus

Archestra is an open source, self-hosted AI platform for deploying and governing agents across an organization. It provides agentic chat for non-developers, apps and skills, shared projects, a server-side agent runtime, MCP orchestration, permission-aware RAG, LLM and MCP proxies, security guardrails, and observability in one platform. Users sign in with SSO, and every tool call runs under that person’s own identity rather than a shared service account. Projects keep chats, files, scheduled tasks, and instructions together, while agents run in sandboxed containers and can start from schedules, emails, or webhooks. MCP servers run in the organization’s own Kubernetes environment and move through security-reviewed promotion flows with separate credentials and network policies. Knowledge bases can connect Confluence, Jira, drives, and internal documents while preserving source-system ACLs, so users only retrieve content they are already allowed to access.

Phinite provides shared infrastructure for building, deploying, and governing AI agents across orchestration, security, observability, lifecycle management, and environment promotion — so engineering teams don't rebuild these layers for every new agent use case. Core capabilities: Orchestration for multi-agent systems (agent-to-agent, nested calls) Deep session-level observability: execution timelines, decision variables, tool calls, latency/cost tracking Private Agent Registry for skill discoverability Eval suite for accuracy/safety benchmarking Dev-to-Production workflow with environment promotion Kubernetes-native deployment, VPC-internal deployability SOC 2 Type 2 compliance