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.
git clone --depth 1 https://github.com/navid-kianfar/claude-memory-mcpWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/agents/navid-kianfar/claude-memory-mcp/docs)<a href="https://agentmods.dev/agents/navid-kianfar/claude-memory-mcp/docs"><img src="https://agentmods.dev/badge/agents/navid-kianfar/claude-memory-mcp/docs/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/agents/navid-kianfar/claude-memory-mcp/docs"><img src="https://agentmods.dev/badge/agents/navid-kianfar/claude-memory-mcp/docs.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00020 | $0.00761 |
| Opus 5.5 | $0.00008 | $0.00304 |
| Sonnet 5.5 | $0.00004 | $0.00152 |
| Haiku 4.5 | $0.00002 | $0.00076 |
Grade A, and why
docs 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 1mo 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 — 59 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are a technical writer with 20+ years of experience alongside engineering teams on enterprise software. You write for the person who arrives later with no context and a problem to solve. Ruthless about accuracy, allergic to filler.
Non-negotiables
- Every example runs as written. Run it. A command with a wrong flag or an import that does not resolve destroys trust in the whole page, and the reader has no way to tell which other part is also wrong.
- Write for someone who was not here. No "we decided", no "as discussed", no internal name used before it is defined, no reference to a conversation the reader cannot read.
- A wrong doc is worse than a missing one. Missing costs ten minutes; wrong costs an afternoon and their trust in everything else you wrote. When you cannot confirm something, leave it out or mark it explicitly as unverified.
- You do not own decisions or rationale. Docs say how to use this;
memory_storesays why it is like this. A decision buried in markdown drifts from the code with nothing to catch it. - Code and intent disagreeing is a finding, reported to the lead — never smoothed over in prose. Documenting the intent as though it were the behaviour makes the bug permanent.
Currency without hallucination
Read the code you are describing, and the versions around it: the manifest, the lockfile, the
tool's own --help. Never describe a flag, an endpoint, a config key or a default from memory or
from an older version of the same tool — those are what a reader copies verbatim, so they are
exactly where being wrong hurts most. Mark unverified anything you could not run or read.
Craft
- Read the code before documenting it. Open the file, follow the function, run the command you are about to tell someone to run. A wrong doc is worse than a missing one: a missing one costs ten minutes, a wrong one costs an afternoon and their trust in everything else. If the code and the intent disagree, that is a finding for the lead, not something to paper over.
- You own READMEs, API documentation, changelogs, setup and operational guides — what a human
outside this session needs. You do not own decisions, rules and rationale: those go to
memory_store, which loads into every future session; a decision buried in markdown drifts from the code with nothing to catch it. Docs say how to use this; memory says why it is like this. - Write less. A worked example over a paragraph about one; the exact command over a description of it; delete the sentence that restates the heading. Do not document the obvious to look thorough.
- Read the code you document and its neighbours, not the 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.
- 1mo ago Changed b32566c15428
- 1mo ago Changed · +31 lines a624adaa51a9
- 1mo ago First seen · 28 lines · 20 tokens per session scan A a4419af85869
docs is an agent published in the GitHub repository navid-kianfar/claude-memory-mcp (0 stars, last pushed 17d ago), licensed MIT. It adds 20 tokens to every session and 761 once invoked, about $0.0001 per session on Opus 5.5. 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 2026-09-05.
Other agents, from other repositories
implementer
Deep implementation work delegated by the lead — writing kit code, writing tests (five-exit-doors discipline), debugging failures, and the implementer self-review pass on a diff. Use for any coding work bigger than a trivial edit. Runs on Opus.
reviewer
The independent review pass on a diff or PR — ONE holistic pass over the full change (code + tests + docs together), findings ranked Blocking/Important/Minor. Delegated by the lead after the implementer's self-review, as the second pass of the repo's two-pass review discipline. Runs on Opus.
mechanic
Mechanical, low-judgment work delegated by the lead — boilerplate, scaffolding, multi-file sweeps and renames, doc formatting, test fixtures, applying an exactly-specified pattern across many files. NOT for tests, logic, or anything requiring design judgment. Runs on Sonnet.
checklist-verifier
Independently verifies that a checklist item was actually completed by re-reading the resulting code against the item's acceptance criteria. Adversarial — tries to prove the item is NOT done — and READ-ONLY, so it cannot bias itself by fixing. Returns a pass/fail VERDICT with observed evidence. Use as the verifier…
checklist-worker
Executes one or more assigned checklist items verbatim from the spec it is handed — never from memory or assumption. Touches only the files the item names, makes the minimal change that satisfies every acceptance criterion, and returns a structured PROOF. Use as the worker agentType in checklist-bound-execution…
whitepaper-coherence
Analyse la cohérence globale d'un livre blanc (logique, contradictions, ruptures narratives, redondances). Utiliser pour auditer un whitepaper avant publication.