Talkamore
About
A wiki about your life that writes itself. Save from any AI chat, recall it in the next.
Details
- Author
- krane-apps
- Categories
- AI
Jump to
Setup
Install Talkamore in your MCP client (Claude Desktop, Cursor, Windsurf, and others).
Repository: https://github.com/krane-apps/talkamore-mcp
Follow the installation instructions in the repository README, then restart your MCP client.
A wiki about your life that writes itself — exposed as a remoteMCPserver, so the AI you already use can read and write it.
Every day you work things out in ChatGPT, Claude, or Cursor — decisions, plans, things you learned — and lose them when the chat ends. Talkamore is the memory layer that fixes that, built on the pattern fromKarpathy's LLM-wiki gist: instead of machine-written memory fragments, an LLM maintainer files what you say into persistent, cross-linked wiki pages about your people, projects, and themes. Each claim carries a receipt back to the day you said it. You can read your memory like a document, edit it like a document, and the maintainer respects your edits — after you touch a page it only ever appends.
Two guarantees the server enforces in code (both verified by incident replays, 2026-08-09): the result text ofsave_to_wikiis generated from the writes that actually executed — never from the model narrating itself — and when a page is full, displaced content moves to a linked archive page instead of being deleted, with the result saying so ("page full — older notes moved to ..."). If the save response says it was filed, it was filed.
This repo is the public integration reference. The server itself is hosted atapi.talkamore.com— there is nothing to install or run.
Listed on the official MCP registry ascom.talkamore/wiki.
claude mcp add --transport http talkamore https://api.talkamore.com/mcp
Add to~/.cursor/mcp.json, then restart Cursor and approve the sign-in prompt:
{ "mcpServers": { "talkamore": { "url": "https://api.talkamore.com/mcp" } } }
Settings → Connectors → Add custom connector, paste the URL, and approve in the browser.
ChatGPT supports custom MCP connectors through Developer mode, with full read and write tools:
- On the web app: Settings →Security and login→ turn onDeveloper mode
- Settings →Plugins→+→ create a connector withhttps://api.talkamore.com/mcp(auth: OAuth) — ChatGPT runs the browser sign-in flow (client ID metadata documents; no registration step)
- In a chat, pick the connector under the composer'sDeveloper modetool — the wiki tools appear in your conversations
On the free tier, use the capsule import attalkamore.com/wikiinstead: ask ChatGPT what it knows about you, paste the answer, and the maintainer files it into pages.
check my talkamore wiki about the launch plan
The point is the loop: what you work out in one chat is waiting in the next, in any client, and it compounds instead of resetting.
- OAuth: S256 PKCE only, single-use 60s authorization codes, refresh rotation — reuse of a rotated token revokes the whole grant
- Disconnect any client from talkamore.com/wiki; its next call is refused
- Tokens and codes are stored as SHA-256 digests only
- All wiki content is encrypted at rest with per-user keys (envelope encryption)
- Every credential is scoped to one user's wiki; there is no cross-user surface
- Product:talkamore.com
- The wiki surface:talkamore.com/wiki
- Registry entry:com.talkamore/wiki
- The idea this implements:karpathy/llm-wiki
Questions or problems:team@talkamore.com
Local-first agent memory: a plain-Markdown Obsidian vault is the source of truth, with a rebuildable DuckDB index for hybrid BM25 + vector + graph recall.
Persistent memory and semantic search for AI coding assistants across sessions
Give your agent a memory: shared, cited, tenant-isolated knowledge-graph memory for any MCP host. Grounded answers from a local-first June endpoint — abstains rather than guesses.
Decentralized persistent memory for AI agents — encrypted vault storage built on Walrus and Sui.
Persistent memory for AI assistants and coding agents across ChatGPT, Claude, Cursor, and other MCP-compatible tools.
Your portable AI memory vault — memories, skills & configs, shared across every AI tool.
Local Work Model for AI agents that learns from real outcomes.
Adaptive MCP memory system for AI applications. Learns which retrieval strategies work for your data, scores results using cognitive science models, builds a knowledge graph automatically, and validates every parameter change against real query history before adopting it. Patent pending.
Auditable, self-improving knowledge & memory for AI agents over MCP — citation-enforced answers and a replayable why-trace, self-hosted on Postgres.
Turns your task manager into agent memory: hybrid (RRF) retrieval over TickTick or an Obsidian vault via an adapter contract. MCP server + CLI, no vector DB to maintain.
Sign in to leave a review
Use Google, GitHub, or an email account so ratings stay tied to real people.
No reviews posted yet.
