Palim
Measured conclusion: Deep-tested. L2 outcome: tools_ok.
Hosted MCP memory: save sessions/decisions once, search from Claude, Cursor, ChatGPT. EU-hosted FTS.
| Author | io.github.joleschmidt |
|---|---|
| Access | remote |
| Publisher key required | No |
| Measured state | Deep-tested |
| L1 liveness | alive |
| L1 latency | 261 ms |
| Last measured | 2026-10-24 |
Add to your agent
Copy this into your MCP client config:
{
"mcpServers": {
"Palim": {
"url": "https://api.usepalim.com/mcp",
"transport": "streamable-http"
}
}
}
Measured tools (11)
palim_resumeread-onlyContinue the user's most recent Palim thread. Call first in a new chat, no args. Ignore the result if it is unrelated. Project filter is case-insensitive and returns the most recently updated match.
Parameters:
project— string — Optional: only consider threads belonging to this project (case-insensitive; Logisoft ≡ logisoft).tool_source— string — Optional: only consider threads from one client (cursor, claude, codex, chatgpt, ...).within_hours— number — Optional: ignore threads older than this many hours. Omit to always return the latest thread.include_messages— boolean — Optional: append the tail of the transcript when the thread has stored messages. Default false.
Hints: read-only yes · destructive no · idempotent yes · open-world no
palim_checkpointCall after connecting AND again after each meaningful chunk of work. Claude Code: reuse claude-code-<client session id>.
Parameters:
context(required) — string — Handoff another AI can continue from: goal, state, constraints, next steps.tags— array — Optional search tags.title— string — Chat title when visible.origin— string — human (default) or automation (hooks/cron).status— string — Optional status, e.g. laufend or abgeschlossen.project— string — Optional project name.created_at— string — Optional ISO 8601 chat date from the client.key_topics— array — Main topics.open_items— array — Open tasks.session_id— string — Stable chat ID. Reuse it so checkpoints update one row. Claude Code/Codex: claude-code-<client session id>.source_urls— array — Optional URLs.tool_source— string — Client: cursor, claude, chatgpt, … Optional; defaults from the MCP client name, else unknown.continues_from— string — Optional session_id from palim_resume that this chat continues.decisions_made— array — Decisions in this chat.referenced_files— array — Relevant files.resolved_open_items— array — Previously open items that are now done; quote a leading fragment.
Hints: read-only no · destructive no · idempotent yes · open-world no
palim_save_contextLightweight checkpoint: self-contained handoff, no transcript. Reuse session_id in this chat. Proactive after results and decisions, not only once.
Parameters:
context(required) — string — Handoff another AI can continue from: goal, state, constraints, next steps.tags— array — Optional search tags.title— string — Chat title when visible.origin— string — human (default) or automation (hooks/cron).status— string — Optional status, e.g. laufend or abgeschlossen.project— string — Optional project name.created_at— string — Optional ISO 8601 chat date from the client.key_topics— array — Main topics.open_items— array — Open tasks.session_id— string — Stable chat ID. Reuse it so checkpoints update one row. Claude Code/Codex: claude-code-<client session id>.source_urls— array — Optional URLs.tool_source— string — Client: cursor, claude, chatgpt, … Optional; defaults from the MCP client name, else unknown.continues_from— string — Optional session_id from palim_resume that this chat continues.decisions_made— array — Decisions in this chat.referenced_files— array — Relevant files.resolved_open_items— array — Previously open items that are now done; quote a leading fragment.
Hints: read-only no · destructive no · idempotent yes · open-world no
palim_save_sessiondestructiveSave or append a full chat transcript (messages + summary). Exclude children's personal data under 16/higher consent age, HIPAA PHI, PCI payment card data, and adult content; request redaction before saving. Prefer palim_save_context for compact handoffs. For metadata-only or URL updates use palim_update_session_metadata / palim_update_session_references. When appending: reuse session_id + title, fetch existing messages first, send only new ones. Always pass a complete summary and metadata.tool_source.
Parameters:
title(required) — string — Session title. Reuse across saves; else the exact chat title from the client UI.messages(required) — array — Messages to save (role + content). When appending, only unsaved messages.summary— object — Full-conversation summary. Required arrays may be empty.metadata— object — Session metadata. tool_source is required.overwrite— boolean — Replace all existing messages instead of appending. Default: false.created_at— string — Optional ISO 8601 creation date from the AI client (new sessions only).session_id— string — Session id. Reuse across saves in this chat; else from chat URL or a new UUID.
Hints: read-only no · destructive yes · idempotent no · open-world no
palim_get_contextread-onlyGet formatted context for a specific topic by searching and retrieving relevant sessions.
Parameters:
topic(required) — string — Topic to get context for.tags— array — Filter by metadata tags (match any).status— string — Optional filter by metadata.status (e.g. "laufend", "abgeschlossen").project— string — Optional filter by projectmax_sessions— number — Maximum number of sessions to include (default: 3)
Hints: read-only yes · destructive no · idempotent yes · open-world no
palim_get_brainread-onlyDistilled knowledge profile. No args: recent topic index (capped). With topic: full document. Pass limit for a larger or full index.
Parameters:
limit— number — Max topics in the index (most recently updated first). Default 20 when omitted. Pass 0 for every topic.topic— string — Optional topic slug or name. When set, returns the full topic document instead of the index.
Hints: read-only yes · destructive no · idempotent yes · open-world no
palim_searchread-onlySearch saved sessions and memories. Query matches session text or memory content/id. Supports metadata filtering via tags/project/status/tool_source. Omit query for tags-only or project-only session filtering. Project match is case-insensitive.
Parameters:
tags— array — Filter by metadata tags (match any).limit— number — Maximum number of results (default: 5)query— string — Full-text search query. Omit for tags-only or project-only filtering.status— string — Optional filter by metadata.status (e.g. "laufend", "abgeschlossen", "referenz", "archiviert").project— string — Optional case-insensitive project filter (Logisoft ≡ logisoft). Allowed without query or tags.tool_source— string — Optional explicit filter by tool source (cursor, claude, chatgpt, web, etc.).
Hints: read-only yes · destructive no · idempotent yes · open-world no
palim_add_memoryStore a memory, fact, decision, or preference for later retrieval.
Parameters:
content(required) — string — Content of the memorytype(required) — string — Type of memorytags— array — Tags for categorizationmetadata— object — Additional metadatasource_session_id— string — Optional reference to source session
Hints: read-only no · destructive no · idempotent no · open-world no
palim_delete_memorydestructivePermanently delete one stored memory by ID. Use for user-requested cleanup or deletion; this cannot be undone.
Parameters:
memory_id(required) — string — ID of the memory to permanently delete
Hints: read-only no · destructive yes · idempotent yes · open-world no
palim_delete_sessiondestructiveDelete a chat session. Soft delete by default (reversible). Use hard_delete=true for permanent deletion.
Parameters:
session_id(required) — string — The session ID to deletehard_delete— boolean — If true, permanently deletes all data. Default: false (soft delete).
Hints: read-only no · destructive yes · idempotent yes · open-world no
palim_helpread-onlyShow an overview of all available Palim tools and usage instructions.
Parameters:
category— string — Filter by category. Default: all.
Hints: read-only yes · destructive no · idempotent yes · open-world no
Measured by MCPYP probing. Untested fields are omitted, never estimated. Listing is always free; MCPYP never proxies calls or stores user tokens.