memory

A long-term memory system for AI agents. It keeps a small profile in every conversation and stores dated past events in an archive that the agent can search when needed.

In plain words
What is it for?
Use it to retain user preferences, active projects, decisions, open questions, and searchable records of earlier conversations.
Why use it?
It helps an agent remember important preferences and ongoing work without placing every past detail into every prompt. This reduces repeated explanations and keeps routine conversations smaller.

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/marlburrow/hivekeep/memory
Clone the repo
git clone --depth 1 https://github.com/MarlBurroW/hivekeep
Per session 8 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,140 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 $0.00008 $0.01140
Opus 5 $0.00004 $0.00570
Sonnet 5 $0.00002 $0.00228
Haiku 4.5 $0.00001 $0.00114

Measured yesterday against content hash 6336de5715cd, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

memory 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.

docs-site/src/content/docs/agents/memory.md · 88 lines

How it starts

The opening of the file, as written. The whole thing — 88 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Hivekeep gives every Agent persistent long-term memory, in two layers:

Layer What it holds How the Agent sees it
Profile What the Agent knows: current state, standing preferences, active work Always present in its context
Archive What happened: dated events, past details, one-off facts Searched on demand with recall

The split is what keeps memory both reliable and cheap. The profile is small and always there, so the Agent never has to get lucky with a search to know who you are and what you are working on. The archive is unbounded and costs nothing until queried, so it can keep everything else.

The profile

A short markdown document (default budget: 1500 tokens) injected into every prompt. It has conventional sections: Pinned, Active projects, Preferences & conventions, Key decisions, Open threads.

It is maintained three ways:

  • Automatically, during compaction: the maintenance pass rewrites it, folding in what is new and dropping what is resolved.
  • By the Agent, with edit_profile, when you tell it something durable and it should not wait for the next compaction.
  • By you, in the Agent's Memory tab: edit the markdown directly, watch the token count, or regenerate the whole document from the archive.

Pinned entries

Anything under ## Pinned is copied verbatim by every automatic rewrite and never edited or dropped. Use it for instructions you want followed forever ("always write GitHub issues in English"). Both edit_profile(..., pin: true) and the editor can put entries there.

The archive

Individual memories, saved automatically during compaction or explicitly with memorize. Each carries a category, an optional subject, an importance score, and a source context describing where it came from (e.g. "While discussing weekend plans, user mentioned...").

Category Use case
fact Objective information (names, dates, technical details)
preference User preferences and habits
decision Decisions that were made and their rationale
knowledge Learned domain knowledge

Read the full file on GitHub · 88 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 · 88 lines · 8 tokens per session scan A 6336de5715cd

Subscribe to this mod's changes

memory is an agent published in the GitHub repository MarlBurroW/hivekeep (51 stars, last pushed 2d ago), licensed MIT. It adds 8 tokens to every session and 1,140 once invoked, about $0.0000 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.

Related

Other agents, from other repositories

planner

복잡한 기능·아키텍처 변경·멀티스텝 리팩토링 구현 계획 전문. 요구사항 인터뷰 → 코드베이스 조사 → 3-6단계 plan.md 생성 + 인수 기준 포함. NEVER 구현. Use proactively when "구현 계획", "설계해줘", "어떻게 만들지", "spec 작성"처럼 코드 작성 전 계획이 필요한 시점. 발산 아이디어가 필요하면 dev-brainstormer 먼저, 아키텍처 판단은 architect 사용.

sangrokjung/claude-forge · 112 tokens

adversarial-reviewer

Independent read-only checker for behavioural changes. Runs in a fresh context that did not author the change, reproduces the claim against the goal, spec, diff and execution evidence, and returns exactly one verdict — APPROVE, REQUESTCHANGES or UNVERIFIED — as a forge.review/v1 envelope. MUST BE USED before claiming…

sangrokjung/claude-forge · 102 tokens

refactor-cleaner

데드 코드·미사용 exports·의존성 제거, 중복 통합 전문. knip/depcheck/ts-prune 감지 → Grep 참조 검증 → 안전 제거. 피처 브랜치에서만 동작. Use proactively when "데드 코드", "미사용 코드", "정리해줘", "클린업", "리팩토링" 요청 시. 빌드 에러 수정은 build-error-resolver, 새 기능은 tdd-guide 사용.

sangrokjung/claude-forge · 109 tokens

systematic-debugger

Specialist for bugs that reproduce but whose root cause is unknown. Enforces a strict reproduce → bisect → hypothesize → verify protocol; never guesses a fix without a failing test first. Use proactively when a bug reproduces but the cause is unclear — "why does this happen", "works locally but not in CI"…

sangrokjung/claude-forge · 170 tokens

defining-agents

Agent "defining-agents" from sixb-ai/sixb, covering defining agents, config, the model, instructions vs agent skills and tools.

sixb-ai/sixb · 0 tokens

overview

An agent is a conversational assistant you define alongside your ontology. It calls a language model, selected worker tools, sandboxed read and bash, and an authorized Sixb API.

sixb-ai/sixb · 0 tokens