v0.1.0 — first working release

Self-hostable context memory for LLMs, served as a remote MCP server on Cloudflare Workers.
Curate Markdown about yourself, your work, and your preferences — Claude pulls it automatically
instead of you re-pasting it every session.

What works

• Read tools: get_context, list_sections, get_meta — pull curated context, gated by READ_TOKEN
• Write tools: append_journal, list_journal — agents leave notes, gated by WRITE_TOKEN; never touches curated content
• Manual promotion gate: npm run promote reviews the journal, you hand-pick what becomes permanent
• Dual auth: plain Bearer token (Claude Code CLI) and OAuth via DCR (claude.ai connector UI), both converging on /mcp
• OAuth login gate: /authorize page requires your READ_TOKEN before issuing any grant — no anonymous token minting
• 100/100 tests passing

Verified live

Connected and exercised end-to-end against a deployed Worker — both the Claude Code CLI bearer-token
path and the claude.ai OAuth connector path, KV bindings live, journal round-trip confirmed.

Docs

Full setup in README.md.
Threat model in SECURITY.md.
Live landing page: https://dkritarth.github.io/context-kernel/