Palim

Hosted MCP memory server: save sessions and decisions once, search them from Claude, Cursor, ChatGPT, and other MCP clients.

От сообществаРаботаетБез входаГлобальныйБесплатноМожет изменять данные

Что умеет

Palim Resume: Continue where the user left off. Call this as the FIRST action in a new chat, with no arguments, before answering — it returns the most recently updated Palim thread already formatted a
Palim Save Context: Default lightweight Palim checkpoint. Save a self-contained summary of the current chat so another AI tool can continue with the same knowledge. Use proactively after every substan
Palim Save Session: Save or append to a chat session with messages and summary. For metadata-only changes (tags, status, project), use palim_update_session_metadata. For URL/reference updates, use pal

Какие данные видит

Нужен ли аккаунт

Не нужен: сервер работает без входа

Hosted MCP memory server: save sessions and decisions once, search them from Claude, Cursor, ChatGPT, and other MCP clients. EU-hosted. Explicit save + PostgreSQL full-text search (not embeddings). Setup: https://api.usepalim.com/setup — Docs: https://usepalim.com/en/docs/ — Examples: https://github.com/joleschmidt/palim-mcp-examples

Список инструментов сервера (45)

Технические названия из tools/list. Нужны только разработчикам.

palim_resume	Continue where the user left off. Call this as the FIRST action in a new chat, with no arguments, before answering — it returns the most recently updated Palim thread already formatted as a handoff (goal, state, decisions, open items), the automatically recorded last turn where one exists, and a short list of other recent threads. Use it instead of asking the user to summarise a previous chat, and instead of palim_get_context when there is no specific topic to search for. Short threads are returned too: brevity is not irrelevance. If the returned thread turns out to be unrelated to what the user actually asks, ignore it silently.
palim_save_context	Default lightweight Palim checkpoint. Save a self-contained summary of the current chat so another AI tool can continue with the same knowledge. Use proactively after every substantial result, decision, or change of direction; do not wait for the user to ask. This does NOT require or store the message transcript. Reuse session_id from the previous checkpoint in this chat so the same summary is updated instead of creating another session. Use palim_save_session only when the full transcript is actually needed.
palim_save_session	Save or append to a chat session with messages and summary. For metadata-only changes (tags, status, project), use palim_update_session_metadata. For URL/reference updates, use palim_update_session_references. When appending to an existing session: reuse the same session_id and title, call palim_get_session first to check existing messages, then send only new messages. Always include a complete summary reflecting all conversation content. metadata.tool_source is required. If user modeling is enabled, include summary.user_insights as a structured profile patch. Extract ONLY long-term, personally meaningful information: relationships ({name, relation, expertise?, company?, preferences?}), life_phases ({label, status: active\|completed\|planned}), personality ({decision_style?, values?}), preferences ({likes?: string[], dislikes?: string[]}), brand_preferences ({category: string[]}), interests (real hobbies/passions, NOT session topics), goals (life/career goals, NOT session todos), decisions (strategic choices, NOT trivial session decisions), employer, role, location. ANTI-SLOP: Do NOT include filenames, paths, branches, commit hashes, session-specific tasks, generic AI phrases, environment variables, or any info that would be irrelevant in 3 months. Only include fields with clear signals – do not guess.
palim_create_connection_url	Create a URL-based connection for the current user. Returns a full MCP server URL with a short token (e.g. https://.../mcp?token=ctxu_...), plus usage instructions for common clients.
palim_search	Search across all saved sessions. Supports full-text search via query, metadata filtering via tags/project/status/tool_source, or both. Omit query for tags-only filtering.
palim_get_context	Get formatted context for a specific topic by searching and retrieving relevant sessions.
palim_get_session	Retrieve a complete chat session by ID, including all messages and summary.
palim_get_sessions	Retrieve multiple chat sessions by ID in one call. Use after palim_search or palim_search_by_date_range when you need full content for several sessions (e.g. "what did I do last week?"). Prefer this over multiple palim_get_session calls.
palim_show_session	Display a chat session in a readable message-by-message format. Accepts session_id (slug) or UUID. For chats with >20 messages, prompts for display count unless max_messages is set.
palim_get_summary	Get only the summary for a session without loading messages.
palim_list_summaries	List saved session summaries with optional filters. Returns summaries without full message content.
palim_add_memory	Store a memory, fact, decision, or preference for later retrieval.
palim_search_memories	Search through stored memories by query, type, or tags.
palim_update_memory	Update one stored memory by ID. At least one editable field is required; contents, tags, and metadata remain encrypted.
palim_delete_memory	Permanently delete one stored memory by ID. Use for user-requested cleanup or deletion; this cannot be undone.
palim_set_user_context	Set the user-specific context block used for personalized AI guidance.
palim_get_user_context	Get the current user-specific context block.
palim_get_brain	Get the distilled knowledge profile ("Brain"). Without arguments returns the brain index: one line per topic (title, slug, hook, entry count) — cheap to load at conversation start. Pass topic (slug or name) to load one full topic document (decisions, open items, session log). Topics are distilled automatically from saved sessions.
palim_export_brain	Export the entire distilled knowledge profile as an Obsidian-ready markdown vault: an INDEX.md with wiki-links plus one markdown document per topic (with frontmatter), each preceded by a "<!-- file: brain/<slug>.md -->" marker. Use when the user wants to download, back up, or sync their brain into Obsidian, a git repo, or local files.
palim_update_brain_topic	Edit a brain topic: replace its markdown content, rename it, update its index hook, or delete it. Use when the user wants to correct, clean up, or remove distilled knowledge. Keep the section structure ("## Decisions", "## Open Items", "## Session Log") when replacing content.
palim_backfill_brain	One-time catch-up: distill sessions saved before the Brain existed (or before it was on) into the knowledge profile, using each session's already-stored summary. Safe to run more than once — already-distilled sessions are deduped, not duplicated. Use when a user with existing saved sessions asks why palim_get_brain / the dashboard Brain tab looks empty despite having saved sessions.
palim_add_user_rule	Add a user rule (workflow/delegation/hints) for personalized guidance.
palim_list_user_rules	List all user rules for the current user.
palim_update_user_rule	Update an existing user rule by rule_id.
palim_delete_user_rule	Delete a user rule by rule_id.
palim_get_user_profile	Get the inferred user profile (if initialized).
palim_update_user_profile	Update profile settings (consent/refinement) and optional profile patch. Use replace_fields to overwrite arrays/objects instead of merging (e.g. clean up interests, update location). Use delete_fields to remove obsolete fields entirely.
palim_delete_session	Delete a chat session. Soft delete by default (reversible). Use hard_delete=true for permanent deletion.
palim_register	Register a new Palim account and receive an API key. Does not require authentication.
palim_login	Login with existing credentials and create a new API key. Does not require authentication.
palim_logout	Sign out from Palim. Deletes API keys and revokes OAuth sessions.
palim_browse_sessions	Browse sessions (machine-readable JSON). For UI/browser. Use palim_list_sessions for human-readable output.
palim_list_sessions	List saved chat sessions with optional filters (tool_source, tags, status, date range).
palim_get_stats	Get statistics about saved chats: total sessions, sessions by tool, recent activity.
palim_help	Show an overview of all available Palim tools and usage instructions.
palim_update_session_metadata	Update metadata of an existing session (title, tags, status, project, created_at, custom fields). Use when user says "change date of chat X to Y". For URL/reference updates, use palim_update_session_references instead.
palim_update_session_references	Add or update source/reference URLs for an existing session without re-saving messages. Include all distinct URLs related to the chat (chat URL, source links, inline references).
palim_search_by_date_range	Search sessions by date range + optional tags. Use when user mentions a time period (e.g. "letzte Woche", "last week", "dieser Monat") AND optionally a topic/project. Example: "Was habe ich letzte Woche bei Logisoft gemacht?" → date_from/date_to for last week, tags: ["logisoft"]. Faster than palim_search when time + topic are both relevant.
palim_export_session	Export a session as Markdown, JSON, or plain text. When save_to_file=true, create the file using the returned suggested_filename and content.
palim_create_summary	Create a summary for a session that does not have one yet.
palim_find_similar	Find sessions similar to a given session based on keyword and content similarity.
palim_migrate_encryption	Encrypt all plaintext sessions/messages/summaries using the server encryption key. Use dry_run=true to preview.
palim_backfill_blind_index	Populate the blind index columns (migration 024) from existing plaintext: session tag and title tokens plus brain topic slug/title/hook. Additive — plaintext columns are left untouched; completion is per field, so re-running fills only what is missing. Use dry_run=true to preview.
palim_clean_session_tags	Remove tag noise left by the old auto-tagger, which split open items and decisions into one tag per word (stopwords like "nach"/"dem", umlaut-stripped fragments like "lschen"). Conservative: only removes tags positively identifiable as tokenizer output and keeps anything uncertain. Use dry_run=true to review the removal list first.
palim_repair_brain_hooks	Repair brain hooks and session-log lines that the old sentence splitter truncated mid-clause (entries ending on "inkl.", "vs." or an ordinal like "am 2."). Re-derives the full text from the source session summaries. Only ever extends an entry that is a strict prefix of the recomputed one, so hand-edited topics are left untouched. Use dry_run=true to review first.