# Pond Recall

> Search your own past sessions (and any additionally granted pond stores) with the pond MCP tools. Use when you need context from an earlier conversation, a past decision, or anything that happened beyond your current context window.

- Skill: `tenequm/pond-recall` (Agent Skill)
- Install (CLI): `npx skillmds@latest add tenequm/pond-recall`
- Raw SKILL.md: https://api.skillmd.com/api/skills/tenequm/pond-recall/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: tenequm (https://skillmd.com/u/tenequm)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/tenequm/pond-recall

---


# Pond recall

If a `pond` MCP server is present, you have lossless search over past sessions. Stores are mounted read-only under `/workspace/extra/pond`; the tools are read-only by design. pond is MCP-native and self-documenting: the tool descriptions (`pond_search`, `pond_get_session`, `pond_get_message`, `pond_sql`) and resources (`schema://pond`, `schema://pond-sql`, `stats://pond`) carry everything you need.

- `pond_search` for concepts and paraphrases; `pond_get_session` to read a whole session (`from="end"` for the latest state) and `pond_get_message` to expand one message with its full tool bodies; `pond_sql` with `contains_tokens` for exact strings, identifiers, or error messages.
- A zero or weak search result is not proof of absence: try another phrasing, a date filter, or a SQL token match before concluding "nothing found".
- Servers named `pond_<store>` are additional stores you were granted; same tools, different corpus.