pi-from-scratch is a coding AI agent for To quickly build a functional AI agent prototype for simple tasks without depending on big model platforms. Pricing: Free. As of 2026-10-22, KanonAgent records 1.3k upvotes. First indexed by KanonAgent on 2026-08-11.
A 600-line TypeScript implementation that lets developers build a basic Pi agent from scratch, ideal for learning and rapid prototyping without relying on large model platforms.
1.3k upvotes
Tracked by Kanon since Aug 11, 2026
Status basis: npm_downloads_7d momentum fell to 0 per 3d from 16 (<= 40% of the prior window)
📈 Timelinewhat changed, and when
- Cooling npm_downloads_7d momentum fell to 0 per 3d from 16 (<= 40% of the prior window)
- Cooling npm_downloads_7d momentum fell to 0 per 3d from 16 (<= 40% of the prior window)
- Cooling npm_downloads_7d momentum fell to -96 per 3d from 128 (<= 40% of the prior window)
- Cooling npm_downloads_7d momentum fell to -102 per 3d from 202 (<= 40% of the prior window)
- Cooling npm_downloads_7d momentum fell to -30 per 3d from 211 (<= 40% of the prior window)
- Accelerating npm_downloads_7d gained 128 in the last 3d vs 53 in the 3d before (ratio 2.4x, threshold 1.5x)
- Cooling npm_downloads_7d momentum fell to -30 per 3d from 211 (<= 40% of the prior window)
- Accelerating npm_downloads_7d gained 202 in the last 3d vs 72 in the 3d before (ratio 2.8x, threshold 1.5x)
- Accelerating npm_downloads_7d gained 211 in the last 3d vs 95 in the 3d before (ratio 2.2x, threshold 1.5x)
- Anomaly daily change 158.0 is 3.9 standard deviations from this entity's own prior mean (-6.7±41.7, baseline excludes this point) — unusual, not necessarily adoption
- Accelerating npm_downloads_7d gained 148 in the last 3d vs 87 in the 3d before (ratio 1.7x, threshold 1.5x)
- Anomaly daily change 158.0 is 3.9 standard deviations from this entity's own prior mean (-6.7±41.7, baseline excludes this point) — unusual, not necessarily adoption
- Cooling npm_downloads_7d momentum fell to 0 per 3d from 87 (<= 40% of the prior window)
- Cooling npm_downloads_7d momentum fell to -48 per 3d from 32 (<= 40% of the prior window)
- Cooling npm_downloads_7d momentum fell to -71 per 3d from 32 (<= 40% of the prior window)
- Accelerating npm_downloads_7d gained 32 in the last 3d after a flat 3d baseline (no prior movement to compare a ratio against)
- Accelerating npm_downloads_7d gained 32 in the last 3d after a flat 3d baseline (no prior movement to compare a ratio against)
- Cooling npm_downloads_7d momentum fell to 0 per 3d from 109 (<= 40% of the prior window)
- Cooling npm_downloads_7d momentum fell to 0 per 3d from 109 (<= 40% of the prior window)
- Anomaly daily change 119.0 is 6.8 standard deviations from this entity's own prior mean (-42.8±23.7, baseline excludes this point) — unusual, not necessarily adoption
- Anomaly daily change 119.0 is 6.8 standard deviations from this entity's own prior mean (-42.8±23.7, baseline excludes this point) — unusual, not necessarily adoption
- first observed Kanon started keeping permanent history for this agent
🔎 Known / Unknown
11 of 23 fields unknown
| Machine-callable | — | unknown | — |
| Open source | 1 | verified | 1 quote(s) |
| Self-hostable | 1 | inferred | 1 quote(s) |
| Bring your own key | 1 | inferred | 1 quote(s) |
| Autonomy level | 2 | inferred | 1 quote(s) |
| Pricing model | — | unknown | — |
| Integrations | — | unknown | — |
“Unknown” means we have not verified it — it is not a “no”. Hard filters never treat unknown as false.
🤖 Agent teardown · coding
Job to be doneTo quickly build a functional AI agent prototype for simple tasks without depending on big model platforms.
AutonomyL2 · tool-calling(evidence: "tool-calling…")
Who it is forDevelopers
Prerequisitesopen source · self-hostable · bring your own API key
PricingFree
Traction · why it is risingGained 447 upvotes on Product Hunt
Why it matters
Its minimal codebase and customizable design make it an excellent starting point and reusable foundation for learning and building AI agents.
Evidence quotesverbatim, from the product’s own materials
“public GitHub repository (README fetched)”— structural
“npm install export NANOPI_API_KEY=your-api-key npm run dev”— readme
“export NANOPI_API_KEY=your-api-key”— readme
“tool-calling”— topics
“需要 Node.js 22 或更高版本，以及一个 OpenAI 兼容 API。”— readme
Signal source: GitHub
Visit official site →
📛 Official badgefor your site / README
Building pi-from-scratch? Pick a style above — the embed code updates live. Deep color control via URL params: bg= / fg= / accent= (hex). It links back to this page.
On mobile tap Share for WeChat / RED (Xiaohongshu) / X; on desktop use Copy text and paste into the app.
FAQ
What is pi-from-scratch?
A 600-line TypeScript implementation that lets developers build a basic Pi agent from scratch, ideal for learning and rapid prototyping without relying on large model platforms.
What does pi-from-scratch do?
To quickly build a functional AI agent prototype for simple tasks without depending on big model platforms.
Why does pi-from-scratch matter?
Its minimal codebase and customizable design make it an excellent starting point and reusable foundation for learning and building AI agents.
How much does pi-from-scratch cost?
Free
Is pi-from-scratch free?
Yes — pi-from-scratch has a free tier. Pricing as stated on its own page: Free
Is pi-from-scratch open source?
Yes — pi-from-scratch is open source.
Can I self-host pi-from-scratch?
Yes — pi-from-scratch can be self-hosted.
How popular is pi-from-scratch?
As tracked by KanonAgent: 1.3k upvotes (first indexed 2026-08-11).
What are the best pi-from-scratch alternatives?
Similar AI agents tracked by KanonAgent: deepseek-harness, gemini-cli, Codédex, LLM Gateway, oh-my-openagent, goose.