What is the best MCP servers, plugins and skill for model evaluation?

By tashan's measured score: Deepeval, Datarobot Agent Skills, Evalview, Fiftyone, Mlflow. the score combines upkeep and freshness, gated by real adoption — every input is public and linked to its source.

Who does model evaluation as part of their job?

This is work our corpus shows people doing that O*NET does not yet have a process step for — agentic tooling post-dates its software survey.

How is the ranking calculated?

From public signal only: npm download volume and publish cadence, GitHub push/release recency, contributor count, registry status, and how often a capability appears in real public agent configs. Nobody can pay to change a score, rank, or listing.

The Index · Model evaluation

Model evaluation

'Evals' — scoring model output and catching regressions. tashan measures 28 capabilities for this work. Ranked on upkeep, freshness and real adoption, every input public. How we measure ›

#	Capability	tashan	Evidence	Health
1	Deepeval primary plugin solid	77	17k ★	active
2	Datarobot Agent Skills primary plugin solid	62	23 ★	active
3	Evalview primary plugin solid	62	124 ★	active
4	Fiftyone primary plugin solid	61	37 ★	active
5	Mlflow primary plugin solid	61	61 ★	active
6	Probabl Skills primary plugin solid	57	74 ★	active
7	Iris primary plugin solid	45	8 ★	active
8	Nnsight primary plugin solid	42	9 ★	active
9	Autoresearch AI Plugin primary plugin solid	40	12 ★	active
10	HuggingFace Skills primary plugin	77	11k ★	active
11	Promptfoo Evals primary plugin	76	24k ★	active
12	Agent Eval Harness · redhat-global-engineering primary plugin	48	5 ★	active
13	Bitfab primary plugin	39	1 ★	active
14	Langsmith primary npm	39	3k/wk	abandoned
15	Claude Performance primary plugin	38	1 ★	active
16	Everdict primary plugin	34	1 ★	active
·	Setup primary skill	not scored yet	11 repos	active
·	Skill Creator primary skill ✓ Anthropic	not scored yet	7 repos	active
19	Langfuse primary plugin thin	70	218 ★	active
20	Nexus Agents · williamzujkowski primary plugin thin	53	16 ★	active
21	Auxiliar supporting npm solid	54	158/wk	active
22	My Pi supporting npm	62	216/wk	active
23	Evals supporting npm	53	224/wk	active
24	Mcpscope supporting npm	52	300/wk	active
25	Trustmodel supporting npm	48	59/wk	active
26	Orizu supporting plugin	40	0 ★	active
27	Prove supporting plugin	31	1 ★	active
28	Plzebo supporting npm thin	51	326/wk	active

Who does this work

O*NET has no process step for this yet — its software occupations were surveyed before agentic tooling existed. We list it because the corpus plainly shows people doing it, and we say so rather than forcing it onto an unrelated step.

Recently changed in Model evaluation

14 changes recorded here in the last 45 days, newest and most serious first.

2026-08-12 Trustmodel — @trustmodel/mcp-server now reaches network
2026-08-12 Trustmodel — @trustmodel/mcp-server now runs a script when it installs
2026-08-16 Langsmith — langsmith-mcp-server now reads as abandoned, was active
2026-08-05 Setup — the tashan score for setup rose to 59.0, from 45.0

This page cannot know what you run. tashan doctor reads your own config and names which of these you have — Pro adds the history behind each, and what to move to.

tashan Pro — $6/mo ›

Other work

Agent configuration Agent development Animation and generated imagery Application development Attribution Audio production Audit Bookkeeping Browser automation CAD modelling Career development Code review Content marketing Contract review Copy editing Copywriting Customer support Dashboards and reporting Data pipelines Data quality Database access Debugging Design critique Document production Exploratory data analysis Financial modelling Hardware design Incident response Infrastructure and deployment Knowledge management Literature review Logistics and fulfilment ML engineering Market analysis Messaging and email Observability PRDs and specs Performance optimisation Process automation Product strategy Project management Prompt engineering Query optimisation Recruiting and hiring Regulatory compliance Retrieval systems Risk assessment SEO Sales pipeline Scientific research Security review Software architecture Statistical modelling Technical documentation Test automation Training material User research Version control Video editing Visual design Web development Web research Web scraping

Occupational data from the O*NET 30.3 Database by the U.S. Department of Labor, Employment and Training Administration, used under CC BY 4.0. tashan consolidated its process steps into the terms practitioners use; O*NET does not endorse this site.

See the full Index ›

tashan Pro$6/mo

14 changes across Model evaluation in the last 45 days. A new advisory, an install script appearing, a maintainer leaving. All listed free above. Pro keeps the series behind each row, so a number today comes with a direction.

The whole series behind any row, back to the first day we measured it
The replacement, named — not just the news that something died
tashan doctor over your own config, on your own machine

Start a 7-day trial › Everything measured on this page stays free.