HQ vs. the alternatives
{/* FACT-CHECK BEFORE WIDE PROMOTION: competitor capabilities below reflect general knowledge as of mid-2026 and were not freshly web-verified. Keep the tone fair. */}
There are good ways to give an agent some context. Most of them are per-repo, per-user, or per-tool. HQ is the layer that makes context team-wide, persistent, and synced — and it works with the others, not instead of them.
At a glance
| Approach | What it is | Scope | Shared & synced across a team | Beyond instructions (skills · workers · policies) |
|---|---|---|---|---|
| CLAUDE.md / AGENTS.md | Markdown instructions committed into a repo | One repo | Via git, per repo | No — text only |
| Cursor / Windsurf rules | IDE rule files & memories | One project, one IDE | Partly; tool-locked | No |
| Copilot instructions | Repo/org custom instructions for Copilot | Copilot only | Org-level, tool-locked | No |
| Raw MCP servers | Connect agents to external tools & data | Per user / per config | No — per-user setup | Tools, not curated memory |
| Agent-memory tools (Mem0, Letta, Zep) | Long-term / conversational memory for agents & apps | Per app or per user | Developer-integrated | Memory, not team skills/policies |
| HQ | A shared context + capability layer over the agents you already use | Whole company, every tool | Yes — synced across the team | Knowledge + skills + workers + policies |
The honest version
These aren’t rivals — HQ is the layer above them.
- CLAUDE.md and MCP are great primitives — HQ uses them under the hood. The gap they leave is everything between repos and tools: company-wide knowledge, shared skills, and rules that travel with the team.
- IDE rules lock you to one tool. HQ’s context follows you across Claude Code, Cursor, and Codex, so switching tools doesn’t reset your agent’s memory.
- Agent-memory tools remember conversations. HQ remembers the company — decisions, people, conventions, policies — and shares that across everyone’s agents.
If a CLAUDE.md is a sticky note on one repo, HQ is the company’s shared memory — synced, governed, and read by every agent automatically. Already using rules files or an MCP server? HQ makes them team-wide and persistent instead of per-person and per-repo.
Try it:
npx create-hq