Enso CLAUDE.md

Project instructions for Enso, a self-hosted personal AI assistant that uses multiple AI agents and information from a user's digital life. They describe its vision and main development context.

In plain words
What is it for?
Guiding development of Enso's data collection, knowledge system, agent workflows, and related features.
Why use it?
They give coding agents the product purpose and architecture needed to work consistently on Enso.

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/proxy2021/enso/claude-md
Clone the repo
git clone --depth 1 https://github.com/Proxy2021/Enso
Per session 11,646 This file is loaded in full into every session.
When invoked 11,646 The same file — it is already loaded in full.
Security scan A 1 finding. 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 $0.11646 $0.11646
Opus 5 $0.05823 $0.05823
Sonnet 5 $0.02329 $0.02329
Haiku 4.5 $0.01165 $0.01165

Measured 2d ago against content hash 7519d00c1cc3, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

Enso CLAUDE.md scanned grade A with 1 finding 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.

Makes network callslowCapability

Not a fault in itself. Listed so you know the mod talks to something, and to what.

Available methods in executor function bodies: `ctx.callTool(name, params)`, `ctx.listDir(path)`, `ctx.readFile(path)`, `ctx.searchFiles(root, name)`, `ctx.fetch(url, opts?)`, `ctx.search(query, opts?)`, `ctx.ask(prompt,
CLAUDE.md · 589 lines

How it starts

The opening of the file, as written. The whole thing — 589 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Enso — CLAUDE.md

For detailed app building guides, API references, template rules, and code examples, see CLAUDE-REFERENCE.md.

Vision

Enso is a deeply personal AI assistant that understands who you are, discovers what you care about, and mobilizes a full team of AI agents to help you make real progress on the goals that matter most.

Every installation is self-hosted, open-source, and fully owned by the user — you own the factory, not just the product.

The Arc: Understand → Focus → Execute

Enso follows a three-phase arc that compounds over time:

1. Understand the user deeply — Enso scans the user's digital life across 12 data sources (Kindle library, YouTube subscriptions, browser history, email, projects, Steam games, movies/TV, photos, Twitter/X, QQ Music, system apps). Each scan ingests content into the Knowledge Cortex — an interlinked wiki of 2000+ pages that builds a semantic map of who the user is, what they know, and what they're drawn to. Cross-source synthesis connects a Kindle book on quantitative finance to an AlphaRank project to a YouTube channel on systematic investing. The system sees patterns the user might not.

2. Identify what matters — From the Cortex, Enso infers Focus Areas — concrete, outcome-oriented goals the user is working toward. Not category labels ("Quantitative Finance") but actionable goals ("Develop AlphaRank into a Market-Beating Quant Tool"). Each focus has clarity levels (emerging → developing → clear), a deeper personal WHY, adjacent pursuits, and evidence grounded in the user's actual data. Focus areas are first-class Cortex citizens — their wiki pages accumulate everything: evaluation briefings, conversation insights, sprint results, and cross-references.

3. Iterate: Evaluate → Discuss → Evolve — Each focus area has a three-step iterative workflow:

  • Evaluate: An orchestration-powered deep study. Multiple AI agents (researcher, codebase analyst, synthesizer) work in parallel to gather web research, analyze project code and sprint history, cross-reference Cortex knowledge, and produce a comprehensive briefing. Progress is visible live in the Evolve tab.
  • Discuss: A clean strategic dialogue with the AI (no tool calls, no app cards — just focused thinking). The AI arrives prepared with the evaluation briefing and all Cortex context, acting as a co-strategist to flesh out the problem space, define success criteria, and build a clear vision.
  • Evolve: The full conversation context feeds into an /Evolve orchestration — a multi-agent sprint where a team of AI agents (Project Leader, Architect, Engineer, QA, and domain specialists) execute on the agreed goals. Each agent is a Claude Code session with role-specific prompts, coordinated through a DAG execution engine. This cycle repeats — each sprint's results feed back into the Cortex, refining the focus and informing the next evaluation.

Read the full file on GitHub · 589 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 · 589 lines · 11,646 tokens per session scan A 7519d00c1cc3

Subscribe to this mod's changes

Enso CLAUDE.md is an instructions file published in the GitHub repository Proxy2021/Enso (5 stars, last pushed 3mo ago), licensed MIT. It adds 11,646 tokens to every session, about $0.0582 per session on Opus 5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

Related

Other instructions, from other repositories