rekal AGENTS.md

rekal AGENTS.md is an instructions file for Codex, OpenCode from janbjorge/rekal. It costs 1,640 tokens per session, scanned A, original, MIT.

Project instructions for rekal, an MCP server that stores persistent memories for language models using full-text search, vector search, and recency in SQLite. SQLite is a database kept in a single file.

In plain words
What is it for?
Use them when editing rekal's MCP tools, database layer, memory features, or tests.
Why use it?
They explain the architecture and coding rules agents must follow when changing the server.

Instructions file for CodexOpenCode

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/janbjorge/rekal/agents-md
Clone the repo
git clone --depth 1 https://github.com/janbjorge/rekal

Made for: Codex, OpenCode.

Wrote 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.

agentmods badge for rekal AGENTS.md

README.md
[![agentmods](https://agentmods.dev/badge/instructions/janbjorge/rekal/agents-md.svg)](https://agentmods.dev/instructions/janbjorge/rekal/agents-md)
Your own site
<a href="https://agentmods.dev/instructions/janbjorge/rekal/agents-md"><img src="https://agentmods.dev/badge/instructions/janbjorge/rekal/agents-md.svg" alt="Measured on agentmods" height="20"></a>
Per session 1,640 This file is loaded in full into every session.
When invoked 1,640 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 $0.01640 $0.01640
Opus 5 $0.00820 $0.00820
Sonnet 5 $0.00328 $0.00328
Haiku 4.5 $0.00164 $0.00164

Measured 3d ago against content hash 55878e5a21ef, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

rekal AGENTS.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 3d 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.

AGENTS.md · 191 lines

How it starts

The opening of the file, as written. The whole thing — 191 lines — stays where its author put it; the contents beside it link to each section on GitHub.

rekal: Agent Instructions

What this project is

rekal is a Model Context Protocol (MCP) server that gives LLMs persistent long-term memory. It uses hybrid search (FTS5 + vector + recency) in a single SQLite file. Python 3.11+, installed via pip install rekal, runs as a stdio MCP server.

Architecture

MCP Client
  │ stdio (JSON-RPC)
  │
  mcp_adapter.py          ← create_server() factory, lifespan (creates/closes DB)
  │
  └── tools/core.py       ← the 3 tools (build_context, store, delete) + register()
                            │
                    sqlite_adapter.py ← SqliteDatabase @dataclass, ALL SQL lives here
                            │
                            ├── SQLite (memories: content, project, tags, timestamps)
                            ├── FTS5 (full-text index, auto-synced via triggers)
                            └── sqlite-vec (vector index)

Key rules

  • SqliteDatabase is a @dataclass holding the aiosqlite.Connection and ALL query methods. Every SQL statement lives here. No SQL in tool files.
  • tools/core.py holds the tool functions as plain async functions plus register(mcp, readonly=...), which attaches them to a server. REKAL_READONLY=1 registers recall only.
  • mcp_adapter.py exposes create_server(): builds the MCPServer with the right instructions, manages lifespan, and calls register.
  • The MCP surface is deliberately minimal (3 tools). Admin operations (health, export, prune, recall) belong in the CLI (rekal/__main__.py), not new MCP tools.
  • No dynamic SQL. SQL strings must be static literals. No f-strings, no string concatenation, no % formatting to build queries. Use subqueries and parameterized ? placeholders instead.

Re-expansion bar

Do not restore conversations, memory links, scratch/TTL, memory types, access counters, or extra MCP tools because they sound useful. Earlier versions carried all of that; benchmarks showed the structure cost tokens (fatter payloads, fatter instructions) without earning them back. Prefer improving injection quality, bench coverage, and write hygiene (skills/CLI) first.

Read the full file on GitHub · 191 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. 3d ago First seen · 191 lines · 1,640 tokens per session scan A 55878e5a21ef

Subscribe to this mod's changes

rekal AGENTS.md is an instructions file published in the GitHub repository janbjorge/rekal (53 stars, last pushed 8d ago), licensed MIT. It adds 1,640 tokens to every session, about $0.0082 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 instructions, from other repositories