Last measured Nov 7, 2026 · 45 real agent questions

The best tools for AI agents, ranked on evidence.

Superintelligence Plugins ranks the tools agents call, category by category. Every ranking says where its order comes from: our own measurement, an independent benchmark with its date, or a catalog that isn't measured yet.

See the search ranking How we measure
0.96
fact@5, search
#1 of 10 search APIs ranked
Synthetic Search APIfull page text, ~47k chars/query · 41/45 fully answeredBest excerpt API: Parallel, 0.92 (0.91 in fast mode at 1/5 the price)
45
real agent questions, graded
10
search APIs ranked; 14 settings measured
11
tool categories
0
affiliate links or sponsors

Web search APIs, measured

Ranked by fact@5: the share of required facts found in the text of the top five results. Full ranking, per-class results and caveats.

Measured by usOur search eval · Nov 7, 2026
10 search APIs ranked by fact@5 on 45 real agent questions, each at its best-scoring setting; top 5 shown. Measured Nov 7, 2026.
Rank Search API fact@5 Context (chars/query) Full answers Answer page first Cost / 1k Latency
1 Synthetic Search API full page text 0.96 Our search eval Share of required facts found in the top 5 results' returned text Nov 7, 2026 47k Whole pages: median characters returned per query Nov 7, 2026 41/45 32/45 In plan Plan-included: in every Synthetic subscription (from $9/mo); 250 searches/hour quota 0.66 s
=2 Jina Search s.jina.ai, whole pages 0.92 Our search eval Share of required facts found in the top 5 results' returned text. Tied: same score to two decimals (0.924). Nov 7, 2026 86k Whole pages: median characters returned per query Nov 7, 2026 36/45 18/45 Token-billed Billed by tokens; 10M free tokens, about 470 searches at its median page size (eval/REPORT.md) 18.6 s
=2 Parallel Search basic mode 0.92 Our search eval Share of required facts found in the top 5 results' returned text. Tied: same score to two decimals (0.917). Nov 7, 2026 5.0k Excerpts: median characters returned per query Nov 7, 2026 35/45 24/45 $5 $5 per 1,000 calls; same $80 free credit 2.7 s
4 Linkup standard 0.87 Our search eval Share of required facts found in the top 5 results' returned text Nov 7, 2026 12k Excerpts: median characters returned per query Nov 7, 2026 33/45 23/45 $5–6 $5–6 per 1,000 calls; $20 free credit 1.5 s
5 Tavily advanced 0.84 Our search eval Share of required facts found in the top 5 results' returned text Nov 7, 2026 8.7k Excerpts: median characters returned per query Nov 7, 2026 29/45 21/45 $10–16 2.8 s

Full ranking of 10 APIs, with their other settings →

Answer page ranked first

Top 5: questions where the first result, read in full, held the whole answer.

  1. Synthetic Search APIfull page text 32/45
  2. Exaauto 28/45
  3. Brave Search APIweb 26/45
  4. Parallel Searchbasic mode 24/45
  5. Linkupstandard 23/45
Search-then-fetch results

Exa /answer, judged

One-paragraph cited answers, checked against a verified key.

  • 37 right
  • 5 partly right
  • 3 wrong
  • of 45 questions

The wrong answers read as fluent and confident. Open the cited page before acting on one. List price $5 per 1,000 calls.

22 answers held every required fact verbatim; the other 23 were read by hand against the answer key.

Read the judgment

11 categories

Each card names what leads and how we know. Big numbers are scores; catalog cards show adoption in small type, never as a score.

01

Web search APIs

Ranked results for a query, returned as text an agent can read.

Measured by us
Leader, fact@5 Synthetic Search API 0.96 full page text · #1 of 10 APIs, 45 real agent questions
02

Answer and research APIs

A question goes in, a cited answer comes out; deep-research runs at the top end.

1 measured · rest catalog
Measured by us so far Exa /answer 37/45 right; 5 partly right, 3 wrong
03

Fetch, scrape and extraction

Turn a URL into clean text, including sites that block bots.

Independent benchmark
Search + fetch, coding tickets Exa (deep) 83.0% openbenchmarks, Sep 15, 2026
04

Library and API docs for coding agents

Current documentation for the library the agent is coding against.

2 measured · rest catalog
Best docs tool, 15 dev and docs questions Context7 0.71 below web search on the same 15 questions (Synthetic 0.97, Tavily 0.88)
05

Code review and code search

Review pull requests with whole-codebase context; search code for agents.

Independent + commissioned
Contested No stable #1 four different leaders of Martian's Code Review Bench in 2026
06

Browser automation

Let an agent drive a real browser: MCP servers, frameworks, hosted browsers, computer use.

Independent benchmark
Leader, Online-Mind2Web Browser Use Cloud 97.0% Steel leaderboard, Oct 11, 2026
07

Memory for agents

Store and recall facts across sessions.

Independent tests · rest catalog
Most adopted, GitHub stars Mem0 68.6k GitHub stars; not a quality score
08

Sandboxes and code execution

Run code an agent wrote, isolated from everything else.

1 independent chart · rest catalog
Not ranked No quality board exists 6 tools, compared on isolation, price and cold start
09

Agent skills and plugin marketplaces

Where agents get packaged skills, plugins and MCP servers.

Catalog only, not yet measured
Most-starred repo in our catalog anthropics/skills 188.8k GitHub stars; not a quality score
10

Observability and evals

Trace what an agent did, what it cost, and score its output.

Catalog only, not yet measured
Not ranked No independent board 5 tools, with published prices
11

Agent payments and auth

Let agents pay per request and get credentials without a human signup.

Catalog only, not yet measured
x402, real 30-day volume $1.43M third-party dashboard, after filtering wash and idle traffic