APIUp to 28% cheaper than official pricesTry the API →
Hermes
Back to News

Vstorm's AgenticOS Turns Company Chat Into a Self-Hosted Agent Workspace

AgenticOS is an Apache-2.0, self-hosted AI agent platform built on Pydantic AI. Ships shared chat, RAG, a real browser, visuals, apps and coding handoff to Claude Code or Codex. Connects 4,700+ MCP servers and 28 model providers including Ollama and vLLM.

Vstorm's AgenticOS Turns Company Chat Into a Self-Hosted Agent Workspace
Source
AlphaSignal
Published
Author
AlphaSignal
Read
1 min read

Takeaways

  • AgenticOS is an Apache-2.0, self-hosted AI agent platform built on Pydantic AI.
  • Ships shared chat, RAG, a real browser, visuals, apps and coding handoff to Claude Code or Codex.
  • Connects 4,700+ MCP servers and 28 model providers including Ollama and vLLM.
  • Governance built in: per-agent budgets, approvals, hash-chained audit log, envelope-encrypted vault.
  • Runs via Docker Compose on 4 vCPU / 8 GB RAM; quickstart installer provisions a working agent.
  • Won the 2026 AAIF Open Agentic AI Innovation Award ; no native SAML, SCIM or MFA yet.

Vstorm has published AgenticOS , an Apache-2.0 platform for building, sharing, and governing AI agents across a company. It places agents and employees in shared conversations while giving organizations control over deployment, model providers, prompts, tools, secrets, and audit data.

Built on Pydantic AI , the project divides work between engineers and subject experts. Engineers register capabilities in typed Python, while colleagues configure agents in a browser by writing instructions, selecting a model, attaching knowledge, and enabling approved tools. Behavioral changes can therefore move through the console when no code change is required.

A complete stack, ready to host

The repository packages an application stack that runs through Docker Compose on company-managed infrastructure. It supports 28 model providers, including access through an existing ChatGPT subscription and local models served by Ollama or vLLM.

LayerImplementation
Agent runtimePydantic AI and pydantic-ai-harness
APIFastAPI
Data and retrievalPostgreSQL with pgvector
Web interfaceNext.js
DeploymentDocker Compose with published images

Tools that extend beyond chat

AgenticOS combines conversational agents with browser automation, coding sessions, generated interfaces, document retrieval, and Model Context Protocol integrations:

  • Shared rooms: Humans and agents can reply in the same conversation. Each message carries an identity, allowing an agent to distinguish participants.
  • Browser automation:

Reporting is indexed from AlphaSignal. Rights remain with the original publisher and cited sources.

Read original report

Sources

  1. 01github.com
  2. 02vstorm-co.github.io
  3. 03ai.pydantic.dev
  4. 04aaif.io