Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add skills/absolutemode/session-recall/setupnpx skills add AbsoluteMode/session-recall --skill setupgit clone --depth 1 https://github.com/AbsoluteMode/session-recallWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00084 | $0.00955 |
| Opus 5 | $0.00042 | $0.00477 |
| Sonnet 5 | $0.00017 | $0.00191 |
| Haiku 4.5 | $0.00008 | $0.00096 |
Grade A, and why
setup scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 94 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Agent-driven setup
You are the installer. Ask the user the three questions below IN CHAT, in the
user's language, then run the commands yourself. Never invent flags — every
command you need is written here. All state lives outside the repo
(~/.local/share/session-recall/), so nothing here touches their project.
0. Install (skip if session-recall --help already works)
pipx install git+https://github.com/AbsoluteMode/session-recall
Plugin (gives the agent recall tools + auto-fresh index): in Claude Code run
/plugin marketplace add AbsoluteMode/session-recall then
/plugin install session-recall; in Cursor add this repository as a plugin
marketplace and run /add-plugin session-recall; other MCP hosts register
session-recall-mcp directly.
1. Three questions (ask in chat, one message)
- Language you mostly work in with your agent (en / zh / ru / other) — picks the bundled embedding model; specialists exist for en and zh, everything else gets the multilingual model.
- Embedder: local or cloud?
- local, zero setup — bundled CPU model, free, nothing leaves the machine, coarser ranking. Also fine: an already-running ollama/lmstudio is picked up automatically.
- cloud (Voyage) — noticeably better ranking + reranker; transcript
surface text is sent to the provider; needs
VOYAGE_API_KEY.
- Solo or team? Solo is the default (nothing is shared, nothing listens). Team = pair with specific colleagues; every answer to them passes a secret scanner and the user's explicit approval.
2. Apply
Language (always):
session-recall setup --lang <en|zh|ru|…> --yes
--yes also runs the first index. Warn first when the history is large
(setup prints the size): on the bundled CPU model months of history can
take a while; that is normal, later runs are incremental.
Cloud embedder: the key must NOT pass through you when avoidable — ask the
user to add export VOYAGE_API_KEY=… (from voyageai.com) to their shell
profile themselves, then re-run session-recall index in a fresh shell.
Never write an API key into any file inside a repository.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2d ago First seen · 94 lines · 84 tokens per session scan A 8acfc296b500
setup is a skill published in the GitHub repository AbsoluteMode/session-recall (12 stars, last pushed 25d ago), licensed MIT. It adds 84 tokens to every session and 955 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
codemark
Manage structural code bookmarks that survive refactoring. Use when the user says "remember this", "bookmark this", "save this location", "load my bookmarks", "what do I have bookmarked", "mark this code", "codemark", or when starting a new session and needing to reload context from a previous session. Also use…
palaia-memory
Check this person's shared memory before answering anything about their own work, and add to it when something is worth keeping. It holds their projects, decisions and reasons, house conventions (naming, commit messages, tests, releases, review and escalation), people and hard-won details, and every AI tool they use…
palaia-messenger
Register this session with the shared directory so other AI sessions and this person's dashboard can see what you are doing, then use it to reach another live session directly instead of relaying everything through the person. Use it at the start of a task, before picking one up, when your work depends on or informs…
palaia-capture
Save what this person tells you into their shared memory, which every AI tool they use can read. Use it the moment they state a decision and the reason for it, a convention, a correction, or a detail that was expensive to work out — including when they mention it in passing on the way to a different question. Save it…
palaia
Local, crash-safe persistent memory for OpenClaw agents. SQLite-backed by default. Semantic search, projects, scopes, auto-capture. After installing or updating, run: palaia doctor --fix to complete setup.
vector-databases
Vector database operations — embed, store, search, and build RAG pipelines.