Observed Signal · Aug 28, 2026 · Technical Release · Source: Aakash Gupta · Impact: 2/5 · Sentiment: Positive
Building a Company OS with Hermes and OpenClaw
This guide and interview explains how to build a Company Operating System (Company OS) using the OpenClaw runtime and the Hermes agent for automated skill generation and self-improvement. Mikhail Shcheglov, CPO at OLX Uzbekistan, has run a version in production for five months and open-sourced the skeleton of his agent on GitHub. The post describes a five-step build process, a metric called "product context coverage," a three-layer memory architecture (raw transcripts, vectors, knowledge graph), and practical prompts and files (USER.md, MEMORY.md, SOUL.md) to scaffold a production-ready Company OS that operates on Slack and Telegram.
Open-sourced agent skeleton and a practical step-by-step guide are useful to product teams and AI practitioners building internal agentic tooling, but the update is narrowly scoped to internal productivity/PM workflows and does not directly shift core AdTech infrastructure.
Track bolt.new Signals & Market Shifts in Real-Time
Polaris7 autonomous intelligence agents track regulatory filings, primary sources, executive changes, and deal flow 24/7. Create your free Explorer workspace to monitor these entities.
Key Takeaways & Evidence Grounding
- Mikhail Shcheglov, CPO at OLX Uzbekistan, has had a Company OS in production for five months.
- Mikhail open-sourced the structural skeleton of his agent on GitHub: https://github.com/mshcheglov1-ux/corporate-waters-ai-agent.
- OpenClaw is used as the base runtime (providing Slack/Telegram gateways, scheduler, persistent identity) and Hermes drives automated skill generation and self-improvement.
- The guide defines a metric called "product context coverage" and a three-layer memory design: raw transcripts (layer 3), vectors (layer 2), and a knowledge graph (layer 1).
Connected Companies & Entities
8 Entities mapped“Bolt.new - Ship AI-powered products 10x faster...”
“Customer.io - Send smarter messages using your product data...”
“Mikhail open-sourced the structural skeleton of the agent that runs his working life at OLX. (https://github.com/mshcheglov1-ux/corporate-wa...”
“OpenClaw is the base runtime because it comes with everything out of the box: Slack and Telegram gateways, scheduler, persistent identity...”
“OpenClaw is the base runtime because it comes with everything out of the box: Slack and Telegram gateways, scheduler, persistent identity...”
“Check out the conversation on Apple, Spotify, and YouTube....”
“Check out the conversation on Apple, Spotify, and YouTube....”
“Check out the conversation on Apple, Spotify, and YouTube....”
Ontology Mapping & Concepts
Related Market Signals & Shifts
Recent verified developments and strategic activity across this market segment.
Building a Team OS with Claude Code
Aakash Gupta and Hannah Stulberg published a paid Substack guide (May 8, 2026) describing how to build a shared "Team OS" using Claude Code. The authors recap examples from DoorDash, Pendo, Google and an independent builder who converged on a common three-layer architecture: a shared context repo (searchable Markdown), agent-accessible knowledge, and natural-language querying for teammates. The piece argues Team OSes make institutional knowledge discoverable to both humans and AI agents, reducing context bottlenecks. The full article includes a technical deep dive, a four-week build plan, usage patterns, adoption playbook and downloadable starter resources for paid subscribers.
Build a Team OS Using Claude Code
This article outlines a practical guide for product managers to build a Team Operating System (Team OS) using Claude Code and a shared repository. It describes a repository architecture (root Claude MD, folder-level CLAUDE.md files, and a .claude/ folder for agents, commands, and skills), an ownership model, and a three-tier context-loading strategy (always-loaded root, folder-level indexes on query, and content loaded on demand) to conserve LLM context window and reduce hallucinations. The piece covers planning workflows (plan mode, lightweight alignment), agent orchestration (temp files, verification prompts), analytics integration (queries, schemas, Snowflake), and operational practices to keep the repo current. Examples, templates, a checklist for feature launches, and recommended daily prompts and automation flywheels are provided.
Hermes Positions Itself as Next-Gen Agent Runtime
A developer analysis compares Hermes Agent and OpenClaw, arguing Hermes shifts the agent model from a local-first personal assistant to a persistent, self-improving agent runtime. Hermes emphasizes curated memory layers (MEMORY.md and USER.md), procedural skills that the agent can create and improve, configurable isolated execution backends (Docker, SSH, Modal, Daytona, Vercel Sandbox), and background sessions accessible via messaging. OpenClaw remains notable for broad channel support and a large community, but the author contends Hermes prioritizes long-term operability, safer execution, and compounding procedural knowledge — traits important for deploying agents as supervised infrastructure rather than ephemeral chatbots.
Track Real-Time Market Signals & Shifts
Set up custom watchlists to receive automated, evidence-grounded executive digests whenever material signals or shifts occur across your tracked landscape.