Methodology
How every ranking on StackCook is worked out. This page is generated from the same code and data that build the rankings, so the rules below are the rules in use.
How a tool is scored
Each use case is split into slots, one per stack layer. Each slot has its own criteria, and each criterion has a weight. A slot's weights always add up to 100.
A tool gets a score from 0 to 10 on every criterion, and each score cites its evidence. The tool's total for the slot is the weighted average of its scores, rounded to 1 decimal. Totals are calculated when the site is built. Nobody types them in.
- A score with no verifying evidence is capped at 6. Community sources alone don't count as verifying.
- A score of 9 or more needs own test or benchmark evidence. Without it, the score is capped at 8.
- A tool missing a score on any weighted criterion gets no total, and the use case stays Coming soon until it's fixed.
- Equal totals are broken by: 1. more verified evidence; 2. lower cost; 3. alphabetical.
Criteria and default weights
Slots for the same kind of layer start from the same criteria and default weights. These are the defaults used by the use cases that are live now.
Framework
| Criterion | Weight |
|---|---|
| Fit for the job | 25 |
| Developer experience | 20 |
| Reliability & production readiness | 20 |
| Openness | 15 |
| Ecosystem & integrations | 10 |
| Momentum | 10 |
Model
| Criterion | Weight |
|---|---|
| Task quality | 30 |
| Cost per typical job run | 20 |
| Latency / speed | 15 |
| Tool-calling & structured-output reliability | 15 |
| Context & modality support | 10 |
| Openness / deployability | 10 |
Knowledge
| Criterion | Weight |
|---|---|
| Retrieval quality & features | 25 |
| Operational simplicity | 20 |
| Cost at MVP scale | 20 |
| Scale & performance | 15 |
| Access control & data governance | 10 |
| Openness | 10 |
Tools and Search / browse
| Criterion | Weight |
|---|---|
| Coverage of required actions | 30 |
| Security model | 25 |
| Reliability / maintenance | 20 |
| Setup effort | 15 |
| Openness / official vs community | 10 |
Evals
| Criterion | Weight |
|---|---|
| Tracing depth & debugging | 25 |
| Eval features | 25 |
| Integration effort | 20 |
| Cost / free tier | 15 |
| Openness / self-host | 15 |
Hosting
| Criterion | Weight |
|---|---|
| Fit for workload | 30 |
| Cost at MVP scale | 20 |
| Deploy DX | 20 |
| Reliability / scaling | 15 |
| Lock-in / portability | 15 |
App builder / IDE, Backend / DB, Auth + payments and Analytics
| Criterion | Weight |
|---|---|
| Fit for the job | 40 |
| Cost | 20 |
| Ease of use | 20 |
| Openness | 20 |
Per-use-case weight changes
A use case can move a criterion's weight by up to 15 points from the default, but only with a written reason. The checks that run on every change reject anything else.
Build a customer-support agent · 3 changes
Model · Task quality: Weight 30 → 15
Support turns are short; policy/tool correctness matters more than peak reasoning.
Model · Cost per typical job run: Weight 20 → 25
Many conversations; cost per resolved ticket dominates.
Model · Tool-calling & structured-output reliability: Weight 15 → 25
Handoff and policy-following via tools is the core of the job.
Build a deep-research agent · 9 changes
Model · Task quality: Weight 30 → 40
Report accuracy depends on persistent multi-step search and reasoning.
Model · Latency / speed: Weight 15 → 5
Research runs are background jobs; minutes are acceptable.
Knowledge · Retrieval quality & features: Weight 25 → 30
Deduping and re-finding fetched sources is the main job.
Knowledge · Access control & data governance: Weight 10 → 5
Sources are mostly public web pages; per-user ACLs matter less.
Evals · Eval features: Weight 25 → 30
Report quality (citations, coverage) has to be graded with datasets and LLM judges.
Evals · Integration effort: Weight 20 → 15
All three integrate with the top frameworks; grading matters more.
Hosting · Fit for workload: Weight 30 → 40
Runs take minutes; long-running background jobs are the core need.
Hosting · Deploy DX: Weight 20 → 15
Deploy speed matters less than run length here.
Hosting · Lock-in / portability: Weight 15 → 10
Run length matters more than portability for this job.
Ship a SaaS MVP in a weekend · 10 changes
App builder / IDE · Ease of use: Weight 20 → 30
Weekend timeline: getting auth, data and payments working fast matters most.
App builder / IDE · Openness: Weight 20 → 10
Lock-in is cheaper to fix after the MVP proves demand.
Backend / DB · Ease of use: Weight 20 → 30
Weekend timeline: getting auth, data and payments working fast matters most.
Backend / DB · Openness: Weight 20 → 10
Lock-in is cheaper to fix after the MVP proves demand.
Auth + payments · Ease of use: Weight 20 → 30
Weekend timeline: getting auth, data and payments working fast matters most.
Auth + payments · Openness: Weight 20 → 10
Lock-in is cheaper to fix after the MVP proves demand.
Hosting · Deploy DX: Weight 20 → 30
Weekend timeline: a one-step deploy matters most.
Hosting · Reliability / scaling: Weight 15 → 5
MVP traffic is small; scale matters later.
Analytics · Ease of use: Weight 20 → 30
Weekend timeline: getting auth, data and payments working fast matters most.
Analytics · Openness: Weight 20 → 10
Lock-in is cheaper to fix after the MVP proves demand.
Evidence
Every score links to its sources. These are the kinds of evidence we use, and how many distinct items of each kind the live rankings cite.
Own test · 0 cited
A test we ran ourselves, with the setup written down. Verifies a score, and can back a score of 9 or more.
Benchmark · 10 cited
A public benchmark or leaderboard result. Verifies a score, and can back a score of 9 or more.
Official docs · 104 cited
The vendor's own documentation, pricing or terms. Verifies a score.
Real build · 10 cited
A public write-up of someone shipping this kind of project. Verifies a score.
Community · 3 cited
Forum posts, social posts and other second-hand reports. Doesn't verify a score on its own: scores backed only by it are capped at 6.
When evidence goes stale
Evidence about model behaviour, pricing or cost is flagged for review 90 days after we last checked it. Everything else is flagged after 180 days. The checks that run on every change list stale items so they get re-checked.
Live, early and published
A use case gets a full page only when it is marked Published or Early and every slot passes all of these checks:
- The slot has at least one real ranked tool.
- No pick is a stand-in still waiting for a real tool.
- Every pick cites at least one evidence item.
- Every pick has a numeric score on every weighted criterion, so it has a total.
Anything else is shown as Coming soon, with no picks and no costs. We never fill gaps with made-up data.
Early means the rankings are sourced but not yet hands-on tested or backed by enough real builds. Published also needs at least 3 real-build write-ups and no stand-in picks.
Right now 3 use cases are live: 0 published, 3 early. See all use cases.
No-pay-for-rank policy
Sponsored never affects rank. Rankings are never sold. Payment, affiliate commission, sponsorship or free credits never change scores, weights or order. The scoring code doesn't read sponsor data at all, and an automated test checks that every ranking is identical with and without sponsors.
Paid placements are labelled Sponsored and sit outside the ranking tables, with at most 3 per use-case page. Affiliate links are marked “(affiliate)” where they appear, and we may earn a commission from them. They are never used as a tie-break. See Sponsor.
Spotted something wrong? Send a correction with evidence.