ToolCairnHealth Tiers
Not all tools are equally maintained. ToolCairn computes a maintenance score for every tool in the graph, then maps it to one of four health tiers that agents can use to make informed recommendations.
The Four Tiers
Excellent
80–100Actively maintained with frequent commits, responsive issue triage, regular releases, and a healthy contributor base. Safe bet for production.
Good
60–79Well-maintained with solid activity. May have slightly slower release cycles or smaller contributor pools, but still reliable.
Fair
40–59Showing signs of reduced activity. May have long gaps between commits, slow issue response, or infrequent releases. Use with awareness.
Poor
0–39Minimal recent activity. May be abandoned, archived, or in maintenance-only mode. Consider alternatives unless the tool is uniquely suited.
Scoring Formula
The maintenance score is a weighted composite of five GitHub-derived signals. Each signal is normalized to a 0–1 range before weighting:
maintenance_score = (
stars_normalized × 0.15 +
commit_frequency × 0.30 +
issue_response_time × 0.20 +
release_cadence × 0.20 +
contributor_count_norm × 0.15
) × 100
Where each factor is normalized to 0–1 range:
stars_normalized = log10(stars + 1) / log10(max_stars + 1)
commit_frequency = min(commits_last_90d / 50, 1.0)
issue_response_time = max(1 - (avg_days_to_respond / 30), 0)
release_cadence = min(releases_last_year / 12, 1.0)
contributor_count_norm = min(contributors / 100, 1.0)Health Signals
Each signal captures a different dimension of project health:
| Signal | What it measures |
|---|---|
| Commit cadence | Recent commit activity — a proxy for whether development is alive. |
| Issue responsiveness | How quickly maintainers respond to new issues. |
| Release rhythm | How often the project ships releases — regular shipping signals stability. |
| Community size | Adoption indicators (stars, downloads), softened against outliers. |
| Contributor diversity | Active contributor count — bus-factor and community health. |
Each signal is one input among several; the exact weighting is a tuning surface that evolves as we learn from outcome reports. Compare tiers across tools, not raw numbers.
How Health Affects Search
Health tiers influence the search pipeline in two ways:
- Relevance rerank — Tools with higher health scores get a small boost during the rerank pass, reflecting that well-maintained tools tend to make better recommendations.
- Final selection — When two candidates are otherwise close, a clear health-tier gap (stable vs. emerging, slowing vs. active) can flip the result, or surface both as a two-option choice for the agent to pick from.
Refresh Cadence
Health scores are recalculated by the GitHub indexer on a rolling basis. Each tool is refreshed at least once every 7 days, with high-traffic tools refreshed more frequently. The updated_at property on each Tool node indicates when its health data was last recalculated.