fundscore — Lighthouse for repos

Deterministic investor-readiness scoring for GitHub repositories. A no-LLM CLI tool and GitHub Action that scores your repo 0-100 across three dimensions — artifacts, business viability, and quality — and tells you exactly what to fix and by how much.

⚠️ Asterisk: The score is a snapshot of repo-readiness signals, not a business valuation. It reflects what your repo communicates, not what your business is. The value compounds over time — score history is a track record that can't be backdated.

What it does

fundscore ✅ PASS
  Overall  : 63.63/100
  Artifacts: 67.4/100  (13/18 checks)
  Business : 57.1/100  (4/7 checks)
  Quality  : 64.0/100
  Round    : Seed → 69.33/100  (check size: $1M - $3M)

  Top fixes (by score impact):
    +6.6 pts  Funding or roadmap document exists (FUNDING.md / ROADMAP.md)
    +5.1 pts  Market / comparables document exists
    +4.4 pts  Risks / honest assessment document exists

Three dimensions, scored 0-100:

| Dimension | What it measures | Weight | |-----------|-----------------|--------| | Artifacts | Does your repo have the docs investors expect? (18 checks: README, FUNDING, ROADMAP, RISKS, LICENSE, tests, CI, security, changelog, contributing, architecture, git activity, contributors, deployed URL, etc.) | 50% | | Business Viability | Does your repo communicate the signals investors look for? (7 checks: monetization clarity, recession resilience, pricing power, tech-enabled margins, contingency depth, market evidence, traction evidence) | 30% | | Quality | Are your docs readable, specific, well-structured, and consistent? (5 heuristic dimensions, no LLM) | 20% |

Plus a Round-Specific Report: auto-infers your funding round (pre-seed / seed / series-a / grant) and shows what investors expect at that round, what you have, and what's missing.

Quick start

# Install
npm install -g .

# Score your repo
fundscore

# Full markdown report
fundscore --md

# JSON output (for CI / programmatic use)
fundscore --json

# Score a specific repo
fundscore /path/to/repo

# Fail CI if score below threshold
fundscore --fail-below 50

Commands

fundscore                    # Score current repo (summary output)
fundscore --md               # Full markdown report
fundscore --json             # JSON output
fundscore --fail-below 50    # Exit 1 if score < 50

fundscore fix                # Show scaffold plan for missing docs
fundscore fix --apply        # Create missing doc templates
fundscore fix --apply --force  # Overwrite existing docs

fundscore history            # Show score over time
fundscore history --save     # Save current score as a snapshot

fundscore badge              # Output SVG badge
fundscore badge --embed      # Output markdown badge snippet
fundscore badge --save       # Save SVG badge to repo

fundscore mcp                # Start MCP server (stdio) for AI agent integration

The three dimensions

1. Artifacts (18 checks)

Deterministic file-presence and content checks. Each has a weight; score = weighted pass rate × 100.

| Check | Description | Weight | |-------|-------------|--------| | readme-exists | README.md is present | 8 | | readme-oneliner | README has a problem statement / one-liner | 7 | | readme-cta | README has a CTA or contact info | 5 | | readme-demo | README has a demo link or screenshot | 4 | | deployed-url | README mentions a live deployed product URL | 6 | | funding-or-roadmap | FUNDING.md or ROADMAP.md exists | 9 | | market-comps | COMPARABLES.md or market section in README | 7 | | risks-honest | RISKS.md or limitations section | 6 | | license | LICENSE file or licensing text | 5 | | tests-or-ci | Tests or CI present (multi-language: JS, Python, Rust, Go, Ruby, Java) | 5 | | security | SECURITY.md or dependency scanning config | 3 | | contact-team | Team or contact info mentioned | 5 | | audience-customer | Target audience / customer identified | 6 | | changelog | CHANGELOG.md or release history | 3 | | contributing | CONTRIBUTING.md exists | 3 | | architecture | ARCHITECTURE.md or docs/ structure | 4 | | git-activity | Commits in last 90 days (not a dead repo) | 5 | | contributor-count | 2+ contributors (team signal) | 4 |

2. Business Viability (7 checks)

Checks whether your repo communicates the signals investors look for. Not "we evaluate your business" — "does your repo say the things investors need to hear?"

| Check | Description | Weight | |-------|-------------|--------| | monetization-clarity | How the business makes money is stated (pricing, revenue model) | 8 | | recession-resilience | Recurring revenue, moat, fixed costs, diversification signals | 6 | | pricing-power | Switching costs, retention, CAC/LTV, upsell signals | 6 | | tech-enabled-margins | Automation, API, scale, AI, self-serve signals | 5 | | contingency-depth | Scenario planning, mitigation, runway, break-even | 4 | | market-evidence | TAM, competitors, positioning, growth rate | 6 | | traction-evidence | Users, revenue, growth metrics, testimonials | 7 |

3. Quality (5 heuristic dimensions)

No LLM, no external API. Pure text analysis.

| Dimension | What it measures | Weight | |-----------|-----------------|--------| | Readability | Flesch Reading Ease approximation | 3 | | Specificity | Concrete numbers, dates, metrics | 3 | | Structure | Headings, lists, code blocks | 2 | | Length | Not too sparse, not too padded | 1 | | Consistency | No contradicting numbers across docs | 1 |

Round-Specific Reports

fundscore auto-infers your funding round and shows what investors expect:

| Round | Check Size | What investors want | |-------|-----------|-------------------| | Pre-Seed | $250k-$500k | Problem, team, early signals of life | | Seed | $1M-$3M | Working product, real market, path to revenue | | Series A | $5M-$15M | Real traction, governance, scale, defensible moat | | Grant | varies | Public benefit, open access, reproducibility |

Each round has required (60% of round score), expected (30%), and bonus (10%) checks. The report shows what you have, what's missing, and your round-specific score.

--fix mode

$ fundscore fix
fundscore fix — scaffold plan (dry run)

  FUNDING.md (fixes: funding-or-roadmap)
  ROADMAP.md (fixes: funding-or-roadmap)
  RISKS.md (fixes: risks-honest)
  SECURITY.md (fixes: security)
  CHANGELOG.md (fixes: changelog)
  CONTRIBUTING.md (fixes: contributing)

  Run `fundscore fix --apply` to create these files.

Generates template files for missing docs. Edit them, re-run fundscore, watch your score go up.

Score history (the moat)

$ fundscore history --save   # Save a snapshot
$ fundscore history          # Show trajectory

fundscore history

  Date                     Score    Artifacts  Business  Quality   Round
  ───────────────────────────────────────────────────────────────────────────
  2026-07-07T11:26:21       51.7       52.6       40.5       66.0   pre-seed
  2026-07-08T09:15:03       58.3       61.2       48.1       66.0   pre-seed
  2026-07-15T14:22:10       67.9       72.4       55.7       72.0   seed

  Trajectory: ↑ +16.2 pts over 3 snapshots

Score history is a track record. A repo that went 42 → 58 → 71 over 6 months tells a story. Competitors can build a better scorer, but they can't backdate your history.

Badge

$ fundscore badge --embed
[![fundscore](https://img.shields.io/badge/fundscore-64%2F100-dfb317)](https://github.com/SunrisesIllNeverSee/fundscore)

Embed in your README. The badge signals investor-readiness, same way a Lighthouse badge signals web quality.

Configuration — .fundscore.yml

# Override the auto-inferred Investor Lens
lens:
  round: seed
  checkSize: "$1.5M"
  teamMode: solo
  naics: "511210"

# Override artifact check weights
weights:
  funding-or-roadmap: 10
  market-comps: 8

# Override business check weights
businessWeights:
  monetization-clarity: 10
  traction-evidence: 8

# Mark checks as required
required:
  readme-exists: true
  funding-or-roadmap: true

# Dimension weights (must sum to a positive number)
scoring:
  artifactsWeight: 0.5
  businessWeight: 0.3
  qualityWeight: 0.2

# Score thresholds (0-100 scale)
thresholds:
  warn: 50
  fail: 30

MCP server (AI agent integration)

fundscore includes a built-in MCP server that exposes the scoring engine as tools for AI agents (Claude Code, Cursor, Windsurf, etc.). This puts fundscore inside the agent workflow — the agent can check your score, suggest fixes, and scaffold missing docs without you leaving the editor.

Setup

Add to your MCP client config:

{
  "mcpServers": {
    "fundscore": {
      "command": "npx",
      "args": ["-y", "fundscore", "mcp"]
    }
  }
}

Or run directly:

fundscore mcp    # starts stdio MCP server

Tools

| Tool | What it does | |------|-------------| | score_repo | Score a repo, return agent-optimized report (dimensions, round analysis, top fixes, missing checks). Auto-saves a snapshot to .fundscore-history/ to build score trajectory passively. | | get_fix_plan | Get scaffold plan for missing docs (read-only). Returns which files to create, what checks they fix, and score deltas. | | apply_fixes | Create template files for missing docs. dryRun=true (default) previews without writing. dryRun=false writes files. force=true overwrites existing. |

How agents use it

The agent can:

1. Check your score after changes — "Your fundscore went from 52 to 61. Here's what's still missing."
2. Suggest fixes proactively — "Your repo has no RISKS.md. Adding one would bring your score from 63 to 67. Want me to scaffold it?"
3. Track trajectory — every score_repo call auto-saves a snapshot, building the score history passively.
4. Give round-specific guidance — "For a seed round, investors expect market evidence. Your README doesn't mention TAM or competitors."

The deterministic advantage

An AI agent calling an LLM-powered scoring tool is circular — the agent is asking an LLM whether the repo looks good to an LLM. An AI agent calling fundscore is asking a deterministic audit tool for a reproducible score. The agent trusts the score because it's the same number every time.

Disable auto-save

# .fundscore.yml
history:
  autoSave: false

GitHub Action

The included workflow (.github/workflows/fundscore.yml) runs on every push and PR:

• Pull requests: Posts a comment with the full markdown report
• Pushes: Uploads fundscore-report.json as a build artifact (30-day retention)

Uses only GITHUB_TOKEN — no additional secrets required.

Design principles

• Deterministic — no LLM, no AI, no external API calls. Same repo, same score, every time.
• Transparent — every check is visible, every weight is configurable, every score delta is shown.
• Honest — the asterisk is prominent. The score reflects what your repo communicates, not what your business is.
• Actionable — --fix mode tells you exactly what to do, and each fix shows its score impact.
• Alive — score history grows with your repo. The trajectory is the product.

License

MIT © SunrisesIllNeverSee