Datoka / Our methodology

Evidence before a badge.

Listing completeness, out of 100

A reproducible information score, separate from verification: name and traceable source 10; meaningful description 10; connection or installation reference 20; documentation or repository 15; explicit free access or numeric tariff 20; protocol 10; source date within 90 days 10; named capabilities 5. Missing criteria earn zero. Colors: below 50 red; 50–69 orange; 70–89 blue; 90–100 green. Sort by listing score to compare completeness across the filtered catalogue. Unknown pricing can never earn the full 100. A pricing page alone is not a numeric tariff. Images, popularity and payment to Datoka earn no points.

The same rules apply to Datoka and external tools. Each detail page shows the points, source date, evaluation date and methodology version. Prices in different units are not compared. A complete description does not prove that a service is safe, reliable or effective; those require separate tests. Publisher claims remain unverified. No paid placement or higher score can be purchased.

Four distinct states

These are separate evidence dimensions, not a guarantee of truth, safety or regulatory compliance. Domain claims and dated observations are displayed separately from the imported record.

Sources, coverage and updates

Datoka reuses its existing product registry. The external index has three directory importers: namespaced MCP server records, A2A package and agent-card declarations, and x402 HTTP resources with declared payment requirements.

A listing is not a unique provider. One provider can have several tools or endpoints; separate directories can describe the same integration. Protocol labels do not create extra listings. Source identity and versions are kept separately, and a matching name or endpoint alone does not justify merging records.

Imports are configured for bounded scheduled work and resumable progress. Configuration is not evidence that a particular run completed. Use dated source records and run evidence to assess freshness; a missing entry in a partial page does not establish withdrawal. Declared deleted or deprecated records are excluded from active results.

What the editorial selection means

The ecosystem references are a separate starting point, not extra rows in the index or a performance ranking. They distinguish providers, APIs, MCP servers, frameworks and protocols. The official documentation review date is shown on each note. Reviewing documentation does not test an operation.

Selection requires a useful category, an understandable publisher identity, traceable official documentation and an original explanation of scope. Metadata with a work-in-progress declaration, a description shorter than 40 characters, an unknown source date, a source date more than 180 days old or missing dedicated documentation is flagged for editorial review. The 180-day review rule is separate from the 90-day freshness criterion of the completeness score.

Incomplete identity/provenance or unsafe links keep a record out of the editorial selection until review. A declared non-active source status also excludes editorial promotion. These flags do not delete catalogue data, prove service failure or establish that a service works. Unknown stays unknown.

The five Datoka local tools link to their existing prerequisites, conditions and result-recovery guides. Their public examples were checked on 27 October 2026. A checked local example is different from a live provider test, ongoing monitoring or independent customer adoption.

Other discovery references

The following sources help compare coverage and find original publisher documentation. They are references for review, not additional Datoka bulk imports. Directory entries, repository licences and rights to data or marks are separate; no blanket reuse permission is inferred.

Search and AI visibility

Canonical links, sitemaps and source-linked editorial HTML help discovery. They do not guarantee indexing, ranking or an AI citation. Compare dated observations over equivalent periods; impressions, clicks, citations and catalogue page counts are different measures. Missing measurements are unknown, not zero. Search access and model-training controls are separate policies.

Methodology and editorial policy reviewed 28 October 2026. This date is not a current health check of every imported service.

Prices, rankings and metrics

Prices are read from explicit publisher pricing fields or directory payment requirements. Protocols come from structured declarations and directory records, not arbitrary keyword matches. Unknown prices stay unknown. Search matches declared text using an FTS index and ranks results by exact name and text relevance. Task guides are editorial groupings, not verified capabilities. Datoka products are identified as such; the sidebar has a dedicated Datoka section. There are no sponsored placements in this release.

No uptime, latency or performance score is invented. The future monitoring system must publish its sampling window and scope before displaying these metrics.

Product images

Logos declared in source records are displayed from supported image hosts. When unavailable, a linked GitHub repository owner avatar may be used and is labelled as such; otherwise initials remain. Images are not proof of identity. External images are requested by your browser without a page referrer. A broken image falls back to initials.

Network safety

The importer contacts only three fixed directory APIs and does not follow redirects. It never visits submitted endpoints, runs packages, calls tools or makes payments. General endpoint checks will require a separate isolated verifier with DNS/IP and redirect protection.