Skip to content

Local-first 🧠 memory for Hermes Agent that lives in your Obsidian vault and remembers project context. Nightly 😴 dream passes turn repeat corrections into confirmed preferences with measurable confidence. Adapters ship for Claude Code, Codex, and OpenClaw, with an MCP server for anything else.

TypeScript

docs: add project README

hermes-vps-agente08df7ef979a
diff --git a/README.md b/README.md
new file mode 100644
index 00000000..286390c3
--- /dev/null
+++ b/README.md
@@ -0,0 +1,63 @@
+# open-second-brain
+
+Open Second Brain is an open-source, plugin-first second brain package for AI agents and humans.
+
+It is designed to give Hermes Agent, Claude Code, OpenAI Codex, and other agentic runtimes a shared, portable way to remember durable project knowledge, append operational event logs, query an Obsidian-compatible vault, and carry the same workflow across machines without locking knowledge into one agent runtime.
+
+Status: experimental v0 design. The first release target is documentation, CLI foundations, skills, and plugin manifests. Deeper runtime integrations and MCP support are planned for later versions.
+
+## Goals
+
+- Provide a filesystem-first second brain that works with Obsidian-compatible Markdown vaults.
+- Keep mutable user data and configuration separate from immutable plugin/package code.
+- Support multiple agent runtimes through adapters instead of a single runtime-specific implementation.
+- Treat daily logs as one backend for an append-only agent event log.
+- Make setup, status checks, redacted config export, and future migrations deterministic through CLI tools.
+- Avoid storing secrets, tokens, credentials, or private connection strings in the vault.
+
+## Non-goals for v0
+
+- No always-on daemon.
+- No mandatory MCP server.
+- No automatic self-rewriting maintenance jobs.
+- No replacement of an existing personal vault.
+- No hidden background writes outside the configured agent-owned area.
+
+## Planned v0 shape
+
+```text
+open-second-brain/
+ docs/
+ idea.md
+ architecture.md
+ roadmap.md
+ skills/
+ open-second-brain/
+ SKILL.md
+ agent-event-log/
+ SKILL.md
+ scripts/
+ asb
+ vault-log
+ plugins/
+ hermes/
+ plugin.yaml
+ __init__.py
+ .claude-plugin/
+ plugin.json
+ .codex-plugin/
+ plugin.json
+```
+
+## Runtime strategy
+
+Open Second Brain is plugin-first, but not plugin-only.
+
+- Plugins provide installation, discovery, lifecycle integration, and runtime adapters.
+- Skills teach agentic runtimes the protocol and safety rules.
+- CLI tools provide deterministic operations that should not depend on model reasoning.
+- MCP can be added later as a shared tool API over the same core.
+
+## License
+
+MIT. See `LICENSE`.