The founder's brain built on Claude + Obsidian. Memora is a founder's persistent memory layer — vault-based, self-organizing, playbook-powered - Includes sales, fundraising, hiring, product, content, strategy and intelligence.
SaferSkills independently audited memora (Agent Skill) and scored it 96/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 1 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 1 flagged
A bulleted imperative like {match} tells the agent to never reveal, disclose, or mention something to the user. Used adversarially it can instruct the agent to hide its tool calls or lie about what it did — stripping the transparency a user relies on to trust the agent.
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
Your company's institutional memory. Persists across every session. Organises itself overnight. Gets sharper the more you feed it.
Memora turns Claude into a founder-grade intelligence layer built on plain-text files you own. Your decisions, your market signals, your GTM rationale, your investor conversations — all of it remembered, cross-linked, and ready to reason from the moment you open a chat.
No re-briefing. No context loss. No starting from scratch.
You re-brief people constantly.
Your CTO. Your first sales hire. A new advisor. A VC on their third meeting. You spend hours reconstructing context that already exists inside your head — because there's nowhere for it to live except there.
Then you open Claude. And you do it again.
Every session starts cold. Every time you close the chat, the context dies. You're the single point of failure for your own company's memory.
Memora fixes this. Your vault holds everything. Claude reads it before answering. You stop re-briefing and start building.
memora/
├── templates/
│ └── CLAUDE.md ← your founder profile, loads every session automatically
├── prompts/
│ ├── 01-founder-profile.md ← Claude interviews you and writes your profile
│ ├── 02-build-venture.md ← spins up a venture folder for any initiative
│ ├── 03-build-playbook.md ← turns any repeated task into a one-phrase command
│ ├── 04-morning-intel.md ← daily brief: filed, flagged, summarised by 7am
│ └── 05-capture-signal.md ← drop in any source, Claude files and cross-links it
└── playbooks/
├── sales/
│ ├── pipeline-review.md ← weekly deal health check, flags stale and at-risk
│ ├── deal-autopsy.md ← diagnose every lost deal into one process change
│ ├── cold-outreach.md ← prospect research → 3 tailored email variants
│ └── customer-expansion.md ← find upsell and expansion before it finds you
├── fundraising/
│ ├── investor-research.md ← pre-meeting brief, right angle, right questions
│ ├── monthly-investor-update.md ← draft your investor update from vault data
│ ├── data-room-checklist.md ← stage-matched due diligence gap analysis
│ └── term-sheet-compare.md ← plain-English term sheet comparison
├── hiring/
│ ├── job-description.md ← role-specific JD grounded in your company context
│ ├── interview-scorecard.md ← 5 dimensions, anchored questions, calibration guide
│ ├── onboarding-brief.md ← first-30-days brief built from the full vault
│ └── reference-check-prep.md ← tailored questions that get past polished references
├── product/
│ ├── prd-from-calls.md ← synthesise discovery calls into a real PRD
│ ├── feature-prioritization.md ← score features against ICP, north star, and GTM
│ ├── release-notes.md ← translate eng tickets into user-facing language
│ └── competitor-feature-gap.md ← analyse any competitor move and respond
├── content/
│ ├── linkedin-from-insight.md ← vault insight → 3 LinkedIn post variants
│ ├── founder-newsletter.md ← weekly debrief → newsletter draft
│ ├── case-study.md ← customer story → full case study + deck-ready short
│ └── content-calendar.md ← 4-week calendar grounded in vault signals
├── strategy/
│ ├── weekly-debrief.md ← founder weekly review + pattern detection
│ ├── quarterly-planning.md ← quarter in review + next quarter goals
│ ├── advisor-brief.md ← pre-call brief with the right ask
│ ├── partnership-eval.md ← evaluate any partnership against 5 dimensions
│ └── board-pack.md ← full board pack from vault data
└── intelligence/
├── industry-signal.md ← filter any market report to what actually matters
└── regulatory-watch.md ← regulatory change → impact assessment + action planRun 01-founder-profile.md once. Claude interviews you across 9 questions — your company stage, ICP, GTM motion, north star metric, active ventures, fundraise status, competitors. It writes all of it into CLAUDE.md.
From that moment, every Claude session opens with full context already loaded. No introduction. No "let me give you some background." You open the chat and go straight to the problem.
Board call on Friday. You say: "run the board-prep playbook."
Claude reads your last 30 days of Outputs and Feedback, pulls your north star metric, surfaces what shipped and what stalled, and hands you a board-ready narrative. You edit for 10 minutes instead of spending 3 hours reconstructing what actually happened.
Same for investor updates. One phrase. Draft from your real metrics. Done.
Every time you read a competitor's pricing page, see a feature drop, or hear something in a podcast — drop it in Inputs.
Claude files it under that competitor, cross-links it to your positioning and ICP pages, and flags: "this changes your differentiation story in the enterprise segment."
Six months in, you have a living competitive map — not a deck that went stale in February.
Your sales motion, your ICP definition, your objection-handling — all of it in the vault.
When you hire your first sales rep, you don't scramble to write a playbook. You say "pull everything we know about our ICP and top objections" and Memora hands them a brief. Onboarding that used to take weeks takes days.
The daily automation runs at 7am. By the time you pick up your phone:
Every Friday: "run the weekly-debrief playbook."
Claude reads all your Feedback and Outputs from the week, checks your north star metric, and writes a founder debrief — what shipped, what stalled (honestly), key signals captured, and the one decision you've been avoiding.
After 4 weeks it starts surfacing patterns across debriefs. The thing you keep saying will happen next week. The metric that keeps slipping for the same reason. You can't unsee it once it names it.
Your investor update format. Your hiring brief structure. Your customer discovery call summary.
Do it once, build a playbook from 03-build-playbook.md. Next time: one phrase, Claude runs the whole workflow from your actual vault data in under a minute.
| What founders do today | What Memora replaces it with |
|---|---|
| Re-briefing Claude every session | CLAUDE.md loads — it already knows you |
| Manually writing board and investor updates | Board-prep and investor-update playbooks |
| Scattered notes across Notion, email, docs | Single vault, cross-linked, fully queryable |
| Forgetting why a decision was made | Every Feedback note is dated and linked to the venture |
| Onboarding new hires from scratch | "Pull everything about our ICP" — done in seconds |
| Competitor research going stale | Living competitive map, updated on each signal |
| Weekly reviews you skip because they take too long | One phrase, 500-word debrief, done in 2 minutes |
| Timeline | What Memora knows |
|---|---|
| Month 1 | Your profile, current bets, active ventures |
| Month 3 | Your decision history, ICP in depth, competitor moves |
| Month 6 | Patterns you didn't notice, 6 months of queryable context, instant onboarding for anyone new |
Most founders carry all of this inside their heads. It dies the moment they close a chat window.
Memora is the system that keeps it alive.
Download from claude.com/download, install, and sign in. You need a paid plan. Open the app and click the Code tab.
Download from obsidian.md (free). Click Create new vault, name it memora, pick a folder, and click Create.
In Obsidian: Settings → Community plugins → Turn on community plugins → Browse → search "Local REST API" → Install → Enable.
Open the plugin settings and copy the API Key.
Modern method (recommended):
claude mcp add --transport http obsidian https://127.0.0.1:27124/mcp/ --header "Authorization: Bearer YOUR-API-KEY"If you get a certificate warning, switch to HTTP:
claude mcp add --transport http obsidian http://127.0.0.1:27123/mcp/ --header "Authorization: Bearer YOUR-API-KEY"Classic method (fallback, requires Node.js or uv):
claude mcp add-json obsidian-vault '{ "command": "uvx", "args": ["mcp-obsidian"], "env": { "OBSIDIAN_API_KEY": "YOUR-API-KEY", "OBSIDIAN_HOST": "127.0.0.1", "OBSIDIAN_PORT": "27124" } }'Test it: In Claude Code, type: list every file in my Obsidian vault. If it reads your notes back, you're connected.
Paste the prompt from prompts/01-founder-profile.md into Claude Code. Claude interviews you once across 9 questions and writes your full profile into CLAUDE.md at the vault root.
From that moment, every session starts with full context.
Each initiative gets its own memory. Use prompts/02-build-venture.md. You get a clean pipeline:
Investor update. Pipeline review. Competitive teardown. Hiring brief.
Use prompts/03-build-playbook.md. Show Claude how you do it once. After that, one phrase runs the whole thing.
claude mcp add google-workspace uvx workspace-mcp --tools calendarAdd Gmail, Slack, or Notion the same way. Grant read-only access wherever you can.
In Claude Desktop: Schedule tab → + New task → Daily 7:00am.
Paste the prompt from prompts/04-morning-intel.md. You wake up to an organised vault and a 3-line brief every morning.
Telling an agent "don't delete this" is a wish, not a setting. Connecting Claude to your vault grants full read/write/delete access — a real data-loss risk.
A boundary enforced by permissions is real. A boundary written in a prompt is not.
A long chat holds context until it collapses: it slows down, burns tokens re-reading itself on every reply, and loses everything when you close the window.
Memora holds context in plain-text files you own. It only pulls what's relevant into each session. One gets slower and more expensive as it grows. The other gets sharper.
This also isn't search over a pile of notes (RAG). It's a structured, pre-reasoned set of pages Claude is meant to reason from — not dig through at query time. That's why answers improve as the vault grows.
The "let the LLM build and maintain the wiki" pattern originates from Andrej Karpathy's LLM Wiki gist. Memora is a founder-focused build of that idea — with the profile schema, venture structure, playbooks, and morning automation wired in.
MIT licensed. Clone it, make it yours.
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.