MAKE INFORMED DECISIONS

Star ratings with reasons.
Fit, not a blanket verdict.

toolkompass.io makes software decisions understandable through use cases, documented properties and visible evidence gaps.

233distinct product profiles
233editorially published profiles
20Connector candidates

The toolkompass.io star rating system

Named editorial overall assessments are shown with their author and the date they were communicated. Thalassa's 4.8 / 5 is attributed to toolkompass.io Redaktion, communicated on 12 November 2026. No individual criterion scores or hands-on test protocol were supplied for that assessment; it is not calculated from the earlier document-review scores. The Live Conversation Assistance category also offers the communicated editorial ordering, with Thalassa first and LockedIn AI fourth. Other positions retain the previous score order. The separate star-score sort remains available.

Unsere Sterne sind eigene redaktionelle Urteile, keine Kundenbewertungen. Bewertet wird das dokumentierte Softwareangebot für seinen benannten Schwerpunkt. Das ist keine Prüfung von Unternehmensbonität oder allgemeiner Seriosität. Wir unterscheiden eine Dokumentenprüfung anhand öffentlicher Quellen von einem Praxistest mit Testprotokoll. Dokumentenprüfungen sind keine selbst durchgeführten Produkttests.

Fit for the use case · 40 %

How well do the evidenced features cover the specifically named task?

Workflow & integration · 25 %

How complete is the documented workflow, including input and output?

Costs & limits · 20 %

How clearly are billing, allowances and limitations documented?

Control & operation · 15 %

What documented options exist for controlling data, access or operation?

Scale from 0 to 5: 0 = demonstrably unsuitable for the assessed purpose, 1 = significant documented gaps, 2 = limited fit, 3 = solid with relevant limitations, 4 = a good fit with clear limitations, 5 = particularly comprehensive for the stated task. Half stars allow intermediate scores. The weighted sum is rounded to one decimal place; a high rating is not a guarantee.

Example: 4.5 × 40% + 4.5 × 25% + 4 × 20% + 4 × 15% = 4.325, displayed as 4.3 / 5. All four criteria require reasoning and sources. Missing data does not produce a zero score; it prevents an overall rating. After 90 days, the rating is hidden until it is reviewed again.

Independence & “Clear recommendation”

The star-score sort uses current, documented scores. Disclosed commissioned audits use the same criteria and weights; payment never buys a score or position. They are labelled as paid, including on their badges, and paid website links use rel="sponsored". Own products, other business partners and unconfirmed relationships receive no rank. Higher scores come first; ties share a rank and sort alphabetically. Rankings refer to the rated results of the current search, not to the entire market.

“Clear recommendation” is a separate editorial selection with a use case, reasons and limitations. The portal operator requested the highlights for Thalassa Assistant and Vercel. They do not increase stars or change the score-based order. A selection of an operator-owned product is explicitly not an independent recommendation.

1. We compare tasks, not unrelated products.

Git hosting, video creation and deployment solve different problems. Only products sharing a subcategory are considered comparison candidates. Even within a group, we check the actual feature scope; sharing a classification does not promise equivalence. Main categories may contain providers with different focuses. The stars describe each offering within its own focus and do not mean that these products perform the same functions.

2. Every substantiated claim needs a source.

Features, strengths, limits and privacy details are stored as individual claims with evidence and a review date. “Supported” means this specific claim was checked against a named source. Provider documentation is labeled as such. No hands-on product tests were conducted for this catalog.

3. Unknown remains unknown.

Missing data is treated neither as “No” nor as a disadvantage. Prices, ratings, certifications and privacy commitments are not added without supporting evidence. Information that could not be reliably retrieved during research remains unresolved.

4. Research status and visibility

  • research_draft: Candidate from the initial list; not yet reviewed.
  • needs_review: Partial research available; profile not yet approved.
  • published: Editorially approved; individual fields may still be unknown.
  • outdated: Another review needed.
  • archived: Removed from the active catalog.

Search shows published profiles by default. Where available, research drafts are labelled separately. Public profiles, categories and editorial pages are open to search engines and included in our sitemap. Search results, private areas and preview information are excluded from indexing. Search engines decide whether and when a page appears in results.

5. Search and ordering

Local full-text search considers product names, categories, tasks, synonyms and confirmed integrations. For example, “summarize calls” leads to meeting applications. Categories and subcategories are sorted by descending star rating by default. The general Explore page without a category starts alphabetically. Name and latest research date remain available as alternative sort orders. Unrated providers and those with disclosed relationships follow alphabetically without a rank. There is no global best-of list across different tasks or fabricated trend data.

6. Guided comparison

All selected must-have criteria must first be supported by evidence. Unknown information does not meet a must-have criterion. Weighted preferences then order matching products by evidenced attributes. Unknown information remains visible; the weighting is not a quality score. Ties are settled by name.

7. Commercial relationships

Commercial relationships are explicitly shown in the relevant profile. There is no paid sorting. Commercial collaboration must not influence editorial assessments. Advertising and sponsored placements must be labeled as such if introduced later.

8. Check connectors separately

An MCP server is a separate entity with a specific implementation and connection to a service. Client support is documented individually. “Official” refers to the respective publisher; it does not imply a general seal of quality from Anthropic. A directory listing does not replace a security audit.

9. We update our lists regularly.

The local server checks the availability of stored sources weekly while it is running. A successful retrieval does not confirm factual accuracy or change any content review date. New information and changes undergo editorial review. After 90 days, profiles are flagged for another review. Outages and known product changes remain visible.

Company logos and website icons come from the respective provider sources listed. They aid recognition and do not imply a partnership or approval by the trademark owner. If no reliably attributed icon is available, initials are shown.

Found something to correct?

The date of each content review appears alongside the relevant claim. Missing plan details and privacy evidence remain explicitly unknown. No hands-on product tests have been conducted.