atomicmemory-python CLAUDE.md

A repository guide for a Python client library that connects to AtomicMemory, a memory service for software agents. It describes the project structure, development tools, and coding rules.

In plain words
What is it for?
Installing dependencies, editing the client package, running tests, formatting and type-checking code, and following repository conventions.
Why use it?
It tells contributors where to find the relevant code and tests and which commands and libraries to use, reducing inconsistent changes.

Instructions file

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 instructions/atomicstrata/atomicmemory-python/claude-md
Clone the repo
git clone --depth 1 https://github.com/atomicstrata/atomicmemory-python
Per session 947 This file is loaded in full into every session.
When invoked 947 The same file — it is already loaded in full.
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 $0.00947 $0.00947
Opus 5 $0.00474 $0.00474
Sonnet 5 $0.00189 $0.00189
Haiku 4.5 $0.00095 $0.00095

Measured 2d ago against content hash d4a38cbe40e5, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

atomicmemory-python CLAUDE.md 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.

CLAUDE.md · 67 lines

How it starts

The opening of the file, as written. The whole thing — 67 lines — stays where its author put it; the contents beside it link to each section on GitHub.

atomicmemory-python — agent rules

Python client SDK for the AtomicMemory memory layer.

Contributor Context

Use this file as the repo-local contributor guide. When behavior depends on another AtomicMemory repo, rely on the published HTTP/package contract and call out any assumption in the PR.

Before changing code, read the relevant local files first:

  • README.md for user-facing installation and usage.
  • pyproject.toml for supported Python versions, dependencies, and tool config.
  • atomicmemory/ for package code and tests/ for expected behavior.

Repository Shape

  • atomicmemory/ — typed Python package.
  • tests/ — unit and integration tests.
  • examples/ — public usage examples.
  • pyproject.toml — package metadata and tool configuration.

Tooling

  • Package management: uv sync (NOT uv pip install).
  • Python interpreter: .venv/bin/python. Never the global python.
  • HTTP: httpx (sync Client, async AsyncClient).
  • Validation: pydantic>=2.7. jsonschema only when validating against external JSON schemas.
  • Lint/format: ruff (line=120). ruff format --check . is part of pre-commit.
  • Types: mypy --strict.
  • Dead code / static health: vulture --min-confidence 90. The whitelist file .vulture_whitelist.py MUST be included in every invocation (it allowlists Protocol-method parameter names and context-manager dunder args). Use the [tool.vulture] config (uv run vulture) or pass it explicitly: uv run vulture atomicmemory tests .vulture_whitelist.py --min-confidence 90. fallow is TypeScript-only and is not part of this repo's verification.
  • Tests: pytest, pytest-asyncio, pytest-mock, respx. Integration tests gated behind -m integration.

Code rules

  • File ≤400 lines (excl. comments/docstrings); function ≤40 lines (excl. catch/finally and docstrings); test files ≤400 lines; tests ≤40 lines.
  • Google-style docstrings (Args:, Returns:, Raises:).
  • Module docstring at the top of every file. For modules that port a specific TS file, include the TS source-of-truth path ("""Port of atomicmemory-sdk/src/.../foo.ts.""").
  • No fallback values; no os.getenv in business code (config flows through Pydantic models constructed at startup).
  • No silent except. No bare except:.
  • Greenfield code — no backwards-compat hacks, no _ private renames for "removed" code.
  • Snake_case for Python attributes; Pydantic Field(alias="apiUrl") aliases preserve TS camelCase wire format.
  • Keep public API behavior aligned with atomicmemory-sdk where both SDKs expose the same concept.
  • Prefer integration tests with a real HTTP path for client behavior; use mocks only for narrow transport errors.
  • Cross-cutting controls live at one chokepoint, enumerated and bypass-tested. When a security/correctness rule must hold for all of a category (every config with an api_url, every input reaching a sink), apply it through one shared helper, not per-surface — and back it with a reflective enumeration test that fails when a new surface lacks it (e.g. test_every_api_url_config_blocks_imds discovers every BaseModel with an api_url field). Tests must exercise the adversarial bypass (the encoding, the key, the header), not just the canonical example, and validate against the downstream consumer's interpretation (the resolver, Postgres, the server), not your own parser. This is the gap that caused AGNT-PY-001's missed EntitiesClientConfig and numeric-IP bypass.

Read the full file on GitHub · 67 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. 2d ago First seen · 67 lines · 947 tokens per session scan A d4a38cbe40e5

Subscribe to this mod's changes

atomicmemory-python CLAUDE.md is an instructions file published in the GitHub repository atomicstrata/atomicmemory-python (11 stars, last pushed 2mo ago), licensed Apache-2.0. It adds 947 tokens to every session, about $0.0047 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.