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.
npx agentmods add instructions/flyset/mymcp/agents-mdgit clone --depth 1 https://github.com/flyset/MyMCPWhat 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 | $0.01666 | $0.01666 |
| Opus 5 | $0.00833 | $0.00833 |
| Sonnet 5 | $0.00333 | $0.00333 |
| Haiku 4.5 | $0.00167 | $0.00167 |
Grade A, and why
MyMCP 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 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.
How it starts
The opening of the file, as written. The whole thing — 138 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
Project Intent
MyMCP is the repository and host package for a local-first MCP server. It currently hosts the Mnemosyne user-governed memory domain in-process.
Principles
- Read before changing, especially
docs/AI_WORKFLOW.md. - Prefer small, explicit MCP tools over broad access.
- Do not add shell-execution or unrestricted filesystem features.
- Do not store secrets, tokens, private keys, or sensitive personal data.
- Keep memory visible, consent-based, and easy to delete.
- Favor simple filesystem-backed schemas before infrastructure complexity.
Guardrails
- ALWAYS read
docs/AI_WORKFLOW.md
Analysis Style
- Analysis is a joint, step-by-step exercise with the user; the agent does not deliver a finished analysis on its own.
- After each short step, present the concrete evidence and stop, so the user can course-correct before the agent draws conclusions.
- Do not write large standalone analysis dumps; keep each step small and reversible so the user can steer the investigation as it unfolds.
Current Scope
This is an early FastAPI MCP host with a Phase 1 kind-qualified runtime seam, a
delivered Phase 2 bundled Mnemosyne extraction, the released MyMCP/mymcp
0.2.0 public host, the prior 0.2.1 Ollama schema-compatibility build, the
0.8.0 MCP principal/session integration over the OAuth Authentication
resource-server foundation, the 0.9.0 configurable protocol-header strictness
build, the 0.10.0 configurable MCP session-lifetime build, and the current
0.10.1 llama.cpp grammar-safe Tool-schema compatibility build.
The bundled Mnemosyne plugin is 0.3.0; memory_recall, memory_list,
memory_remember, and memory_revise are capability contract
1.2.0, while the other four capabilities are 1.1.0. The canonical
Mnemosyne adapter, configuration,
memory domain, and MCP Tool adapters live under mymcp/plugins/mnemosyne/;
bootstrap, runtime, bindings, argument normalization, and host list_tools
remain host-owned. Host configuration schemas 1 through 7 supply immutable
per-process startup intent and loopback packaged-launcher settings; schema 1
retains enabled_plugin_unsupported, schema 2 composes validated external
plugins at startup, and schema 3 adds explicit Authentication intent. Schema 4
adds the operator-bearer-v1 verifier-source selector; schema 5 adds the
oauth-jwt-jwks-v1 issuer selector, immutable startup validation snapshot, and
conditional RFC 9728 protected-resource metadata/challenge; schema 6 adds
configurable MCP protocol-header strictness; schema 7 adds bounded registered-session
lifetimes. The two production
Bearer methods are mutually exclusive, OAuth establishes identity only, and the
literal loopback HTTP resource is a local interoperability exception. Registered
MCP sessions are host-owned, process-local context; Governance policy remains
unimplemented. Repository migration, operational
reconnect/approval checks,
annotated tag creation, and release publication are complete. Startup composition
for operator-installed, operator-trusted external plugins is delivered: MyMCP
validates compatibility and composition, not plugin safety, and plugin changes
require a server restart. Do not add host-managed
installation, dependency environments, hot activation, plugin isolation, or
runtime lifecycle management. Keep changes minimal and protocol-aware.
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.
- yesterday First seen · 138 lines · 1,666 tokens per session scan A 501805117d05
MyMCP AGENTS.md is an instructions file published in the GitHub repository flyset/MyMCP (0 stars, last pushed 28d ago), licensed Apache-2.0. It adds 1,666 tokens to every session, about $0.0083 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.
Other instructions, from other repositories
membro CLAUDE.md
Claude Code instructions for shawn-durrani/membro, covering claude.md, process, safety rules and orientation.
plur CLAUDE.md
Instructions for plur-ai/plur, covering claude.md, what is plur, development, package dependency and version bumps.
remnic AGENTS.md
Instructions for joshuaswarren/remnic, covering remnic - agent guide, architecture boundaries (non-negotiable), upstream references, adapter implementation rules and openclaw compatibility window.
council-lab AGENTS.md
Instructions for loveramarois-byte/council-lab, covering council lab engineering rules, high-risk decision support, change discipline, macos desktop delivery and required review focus.
claude-memory-mcp AGENTS.md
Instructions for WhenMoon-afk/claude-memory-mcp, covering mooncite repository contract, product boundary, fixed architecture, sources of truth and change discipline.
Vibelution AGENTS.md
Instructions for CCDawn/Vibelution, covering vibelution agent rules, 1. identity and priority, 2. global red lines, 3. start and routing and 3.0 每任务先 router.