How we score
One formula for 13 tools. Every criterion, weight and scale is on this page, with each tool's result.
The principle
Each tool gets one score out of 10, calculated by a formula from facts its vendor publishes: the pricing page, plan limits, product pages, documentation and license. Our reviews are built from those pages, read on the date shown next to each price; we do not run the products ourselves.
Two consequences. When a vendor does not state something on the pages we read, the criterion scores 3 out of 10: we cannot credit what is not published. And when a vendor publishes no price, the price criterion scores 2, because an entry price you cannot see cannot be compared.
The ranking never depends on an affiliate program. Tools with no program are scored and ranked like the others, and a program is never an input of the formula. See the disclosure page.
The 11 criteria and their weights
| Criterion | Weight | Scale (score out of 10) |
|---|---|---|
| Entry price | 15% | Cheapest paid plan, in USD a month, billed monthly. $0 → 10, $16 → 9.2, $30 → 8.4, $60 → 7, $200 → 4, $1,000 → 1.5 (straight line between points). No public price: 2. |
| Free plan or trial | 10% | A permanent free plan 10, a free self-hosted edition 8, a time-limited trial 6, one-off credits 5, nothing 1. |
| Billing you can predict | 10% | No platform meter at all (free software you host yourself, where the only bills are your server and your model provider) 10; one unit per whole run 10; one per customer conversation 8; a published rate per feature 6; one unit per step or action 6; credits that follow model prices or task size 4; no public rate 1. |
| Integrations announced | 12% | Number the vendor announces: 100 → 4, 400 → 6, 770 → 7.5, 2,000 → 8.8, 3,500 → 9.4, 9,000 → 10. No count published: 3. |
| Agents and workflows | 12% | Agents and workflows in one product 10, agents only 8, workflows with AI steps only 6. |
| Choice of AI models | 8% | Bring your own key or model 10, several providers to choose from 8, the vendor picks 5. |
| Knowledge base and memory | 6% | Described on the vendor's pages 10; not stated 3. |
| Human approval | 6% | An approval or handoff step described on the vendor's pages 10; not stated 3. |
| Self-hosting | 8% | Free edition with the agent features 10, free edition without them 6, paid enterprise option 5, none 2. |
| Usable without code | 8% | No-code 10, low-code 7, developer tool 4. |
| API and MCP | 5% | API and MCP 10, API only 8, MCP only 7, neither 2. |
| Total | 100% | Weighted average of the 11 criteria |
From the weighted average to the score you see
The weighted average is what we call the raw score. Every tool compared here is a current product a small team could reasonably choose, so we display scores on an editorial scale from 7.5 to 9.6: displayed score = 7.5 + (raw score − 5.5) × 0.571, rounded to one decimal and kept between 7.5 and 9.6. The conversion is a straight line, so it never changes the order.
Labels: Excellent from 9.0, Very good from 8.5, Good from 8.0, Fair below 8.0.
Tie-breaks: no two tools share a score
Tools are ordered by raw score. If two are equal, the lower entry price goes first, then the larger number of integrations announced, then alphabetical order. If two tools would show the same rounded figure, the one lower in that order shows 0.1 less, so each displayed score is unique and the order is always the raw order.
Each tool, criterion by criterion
Scores out of 10 per criterion, then the raw score and the displayed score. Prices read on November 15, 2026.
| Tool | Entry price (15%) | Free plan or trial (10%) | Billing you can predict (10%) | Integrations announced (12%) | Agents and workflows (12%) | Choice of AI models (8%) | Knowledge base and memory (6%) | Human approval (6%) | Self-hosting (8%) | Usable without code (8%) | API and MCP (5%) | Raw score | Score |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
 1. n8nFrom $24/mo · checked Nov 15, 2026 | 8.7 | 8 | 10 | 6.4 | 10 | 10 | 10 | 10 | 10 | 7 | 10 | 8.93 | 9.5/10 |
 2. ActivepiecesFrom $20/mo · checked Nov 15, 2026 | 9 | 10 | 10 | 7.5 | 10 | 10 | 3 | 10 | 6 | 10 | 10 | 8.81 | 9.4/10 |
 3. MakeFrom $16/mo · checked Nov 15, 2026 | 9.2 | 10 | 6 | 9.4 | 10 | 10 | 3 | 10 | 2 | 10 | 10 | 8.35 | 9.1/10 |
 4. LangflowFree, open source · checked Nov 15, 2026 | 10 | 8 | 10 | 3 | 10 | 10 | 10 | 3 | 10 | 4 | 10 | 8.06 | 9.0/10 |
 5. ZapierFrom $29.99/mo · checked Nov 15, 2026 | 8.4 | 10 | 6 | 10 | 10 | 5 | 10 | 3 | 2 | 10 | 10 | 7.90 | 8.9/10 |
 6. StackAIPrice on request · checked Nov 15, 2026 | 2 | 10 | 10 | 6.4 | 10 | 8 | 10 | 10 | 5 | 10 | 8 | 7.71 | 8.8/10 |
 7. DifyFrom $59/mo · checked Nov 15, 2026 | 7 | 10 | 6 | 3 | 10 | 10 | 10 | 3 | 10 | 7 | 8 | 7.55 | 8.7/10 |
 8. LindyFrom $29.99/mo · checked Nov 15, 2026 | 8.4 | 5 | 4 | 9.2 | 8 | 8 | 10 | 10 | 2 | 10 | 7 | 7.37 | 8.6/10 |
 9. GumloopFrom $37/mo · checked Nov 15, 2026 | 8.1 | 6 | 4 | 6 | 8 | 10 | 10 | 10 | 2 | 10 | 10 | 7.36 | 8.5/10 |
 10. BotpressFrom $189/mo · checked Nov 15, 2026 | 4.2 | 10 | 8 | 3 | 8 | 5 | 10 | 10 | 2 | 7 | 8 | 6.47 | 8.1/10 |
 11. Copilot StudioFrom $200/mo · checked Nov 15, 2026 | 4 | 6 | 6 | 3 | 10 | 10 | 10 | 3 | 2 | 7 | 8 | 6.06 | 7.8/10 |
 12. Relevance AIPrice on request · checked Nov 15, 2026 | 2 | 1 | 4 | 8.8 | 8 | 10 | 10 | 10 | 2 | 7 | 10 | 6.04 | 7.7/10 |
 13. VoiceflowPrice on request · checked Nov 15, 2026 | 2 | 6 | 1 | 3 | 10 | 10 | 10 | 10 | 2 | 7 | 10 | 5.78 | 7.6/10 |
Flowise is listed on the site but not scored: it was discontinued (end of life on August 31, 2026). A score would suggest it can still be adopted. See the Flowise review.
What this score does not tell you
- It does not measure how reliable a tool is in production, how good its support is, or how its agents behave on your data. Those need a trial with your own cases.
- It rewards what is published. A vendor with thin public pages scores lower than its product may deserve.
- Integration counts are the vendor's own figures; we do not audit them.
- Prices are US list prices on the date shown. Some vendors price by country.
Updates and corrections
When a vendor changes a price or a plan, we re-read the page, update the data and the date, and the scores are recalculated at the next build. Nothing is edited by hand. To report a change: contact@best-ai-agent-builder.com.