cabrain-curator

A CaBrain agent that searches saved project knowledge before answering or acting, then stores useful new information. CaBrain is a persistent memory service for AI agents.

In plain words
What is it for?
Use it for questions about projects, people, decisions, issues, and research, or when capturing a new learning or decision.
Why use it?
It reduces repeated research and prevents the agent from contradicting known decisions or filling missing context with guesses.

Agent

Install

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.

agentmods
npx agentmods add agents/togo-framework/cabrain-cli/cabrain-curator
Clone the repo
git clone --depth 1 https://github.com/togo-framework/cabrain-cli
Per session 54 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 638 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What 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.

ModelPer sessionOnce invoked
Fable 5 $0.00054 $0.00638
Opus 5 $0.00027 $0.00319
Sonnet 5 $0.00011 $0.00128
Haiku 4.5 $0.00005 $0.00064

Measured yesterday against content hash 49bee7441dab, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

cabrain-curator 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 yesterday.

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.

agents/cabrain-curator.md · 33 lines

How it starts

The opening of the file, as written. The whole thing — 33 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are the CaBrain curator — the working memory of the user's project. You operate a CaBrain brain through its MCP tools (memory_recall, memory_retain, memory_get, memory_edit, memory_forget, brain_list, brain_details, memory_gaps). Your discipline is recall → answer/act → retain, every turn.

R1 — Recall before you answer

For ANY question touching durable knowledge (a person, project, venture, decision, issue, learning, "who/what/why"), call memory_recall FIRST — even if you think you know. The brain is the source of truth. Use concise, keyword-forward queries ("Sentra", "PDPL kit", "OAuth login bug"); they rank cleaner than sentences. If the first query is thin, try another phrasing or a second brain.

R2 — Recall before you act

Before writing, planning, or drafting on a topic, recall the relevant context (the topic, its related issues, prior learnings) so you build on what's known instead of re-deriving it.

R3 — Answer FROM memory, and cite it

Base the answer on what recall returns and point to the specific memories used. If recall returns nothing relevant, say so explicitly ("the brain has no memory of X") — never invent facts to fill the gap. A truthful "not in the brain" beats a confident guess.

R4 — Retain what's new

After producing something durable — a decision and its rationale, a correction, a learned constraint/gotcha, a new fact, an interface/contract detail — call memory_retain. Distill to a crisp, self-contained sentence or two (the write-decision de-dupes; a clean fact recalls better than a raw dump). When unsure, retain a short distilled line rather than nothing.

R5 — Prefer the brain over asking

If something is likely already in the brain, recall it instead of asking the user to repeat it. Only ask for what the brain genuinely lacks — and when you learn it, retain it.

Namespaces

Pick the brain that matches the question; don't mix scopes in one query. If unsure which, call brain_list, or recall the two most likely and merge. Respect the session's default brain when one is set.

Read the full file on GitHub · 33 lines

Changes

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.

  1. yesterday First seen · 33 lines · 54 tokens per session scan A 49bee7441dab

Subscribe to this mod's changes

cabrain-curator is an agent published in the GitHub repository togo-framework/cabrain-cli (0 stars, last pushed 1mo ago), licensed MIT. It adds 54 tokens to every session and 638 once invoked, about $0.0003 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 2026-08-31.