How we grade

Every tool gets a composite score: the simple mean of five 1–5 criteria. No vendor sponsorship, no paid placements — a score reflects what the evidence shows, not who's buying an ad. This page walks through exactly how a grade is built, criterion by criterion, ending with a real scorecard so you can see it composed from evidence.

Composite at a glance

Composite = mean(C1, C2, C3, C4, C5), each scored 1–5. The composite then maps to a plain-English verdict:

CompositeVerdictWhat it means
1.0-1.9Skip
2.0-2.9Situational
3.0-3.9Solid
4.0-5.0Worth it

4.0–5.0 Worth it: strong, recommended on its merits. 3.0–3.9 Solid: good, minor friction or price tradeoffs. 2.0–2.9 Situational: only pays off for specific needs. 1.0–1.9 Skip: weak, or better alternatives exist.

The five criteria

Tap a criterion to expand it.

C1 Value for the price
A 5 looks like
A free tier that covers real use, or flat/transparent/affordable pricing with no per-seat surprises.
A 1 looks like
Quote-only, "contact sales," or opaque pricing.

Backed by: the tool's own pricing page.

C2 Agent Friendliness
A 5 looks like
Agent-native: an official MCP server or a first-class CLI, plus a public API with key-based auth and agent-readable docs.
A 1 looks like
GUI-only, closed, interactive-auth-only.

Backed by: API / developer docs.

A tool's own built-in AI is recorded as a fact, not this score -- C2 is whether an agent can drive the tool, not whether the tool has AI features.

C3 Ease of adoption
A 5 looks like
Self-serve, learnable in a day, no IT/SSO/vendor implementation required.
A 1 looks like
Requires IT, enterprise SSO, or a vendor onboarding process.

Backed by: the signup, plans, and onboarding docs.

C4 Portability & lock-in
A 5 looks like
Open formats (IFC, DWG, PDF, CSV) with easy export -- you can leave whenever you want.
A 1 looks like
Proprietary format, no export -- your data is hostage.

Backed by: import/export and data-export docs.

Higher score = freer.

C5 Interoperability -- fits your stack
A 5 looks like
Native integrations/plugins plus open formats -- joins the Revit / Bluebeam / QuickBooks / Microsoft 365 stack you already run.
A 1 looks like
An island: no integrations, manual export only.

Backed by: the integrations / marketplace page.

The three "connection" axes are distinct. C2 is whether a coding agentcan drive the tool. C4 is whether you can leave the tool. C5 is whether the tooljoins the stack you already run. A tool can score well on one and poorly on another -- they aren't the same question.

Red flags

When present, a red flag caps the criterion it touches -- or, for abandonware, the whole verdict -- regardless of how the other criteria score.

RF1Pricing wallCaps C1 (Value) <= 2.No public pricing -- you have to book a sales call to find out.
RF2Agent-hostileCaps C2 (Agent Friendliness) <= 1.GUI-only, no API/CLI/MCP, interactive-login-only -- an agent can't drive or extract anything.
RF3Hard lock-inCaps C4 (Portability) <= 1.Proprietary formats with no clean way to export your own data.
RF4Fails the existence gateThe tool is not listed at all.A dead or parked site, or no independent source confirms the tool.
RF5AbandonwareCaps the verdict at Situational or below, regardless of composite.In maintenance or sunset -- no meaningful release in a year.
RF6Trains on your dataCaps C4 (Portability) <= 2.The vendor confirms training on customer data with no opt-out (silence isn't this flag -- we mark it unknown).

Provenance & trust

The rule: nothing is listed from memory. Before a tool earns a page, we open its live site and confirm it exists and does what it claims, and we find at least one independent source (a directory listing, press, funding news, a named-customer case study) that isn't the vendor's own marketing -- fail this existence gate and the tool isn't listed at all (RF4). If we can't verify a claim, we drop it rather than guess.

Every claim is checked against the live site. The verdict-driving claims — agent access, sector targeting, pricing — require two independent sources. Each tool entry records the source(s) we checked it against and a lastVerified date; we re-verify pricing every 90 days and everything else every 180 days.

See it graded: a real example

Here's an actual scorecard from the catalog, so you can see a grade composed from evidence rather than take our word for the method.

Autodesk Forma Site Design

Cloud-based AI platform for early-stage urban design, massing, and microclimate analysis.

3.8/5 composite
Solid
CriterionScoreWhy
C1 Value for the price4/5Subscription pricing (verified 2026-07-18); no confirmed free tier on the record.
C2 Agent Friendliness4/5No MCP or CLI, but a documented REST API (OAuth 2.0 via Autodesk Platform Services) and an embedded SDK -- an agent can drive it, just not zero-setup.
C3 Ease of adoption2/5Web-only signup fits the small-team/growing firms it targets, per the catalog's firm-size fit.
C4 Portability & lock-in5/5Exports IFC and glTF (open formats) alongside native RVT -- your data isn't hostage.
C5 Interoperability4/5Native integrations with Revit, Civil 3D, Autodesk Tandem, Esri ArcGIS, Zoneomics -- joins an existing stack instead of sitting apart from it.

Verified against aecmag.com on 2026-07-18.

See the full Autodesk Forma Site Design page →

Browse all 629 tools →