Head to head · Fetching pages · October 2026 research run
Fetch (MCP reference server) vs Hyperresearch
Hyperresearch scores 68.1 (B) on agent readiness against Fetch (MCP reference server)'s 63.8 (B), and leads in 3 of 7 scored categories. Fetch (MCP reference server) leads on reliability, agent ergonomics and payments & pricing. Both do fetching pages.
Best web search APIs for AI agents · All 353 search comparisons
Which one, for what
Fetch (MCP reference server) B
Best for Reading a known URL as Markdown in a local or sandboxed agent, at no cost.
Ahead on
- Reliability, 55 against 35
- Agent ergonomics, 93 against 83
- Payments & pricing, 90 against 60
Also in its favour
- No key needed to call it
- Runs on your own machine
- Open source
Watch for
The address guard doesn't cover DNS rebinding or hosts only a proxy can resolve, per its README
Best for Agents that need cited, quote-checked facts or long research reports and want to keep sources in a searchable vault.
Ahead on
- Schema & documentation, 76 against 70
- Security & auth, 79 against 47
- Maintenance & community, 88 against 72
Also in its favour
- A hosted endpoint, with nothing to install
- Free to start without a card
- No incidents deducted, where Fetch (MCP reference server) loses 6 points for them
Watch for
No status page, and the terms say there is no SLA and no uptime figure
Why the edge
Each category where one leads by five points or more, with the reason behind each side's score from its assessment. The numbers are in the table below; this is what they stand for.
Security & auth Hyperresearch +32
Fetch (MCP reference server) 47
No credentials to leak and nothing to scope, so the middle band (20). GET only. Since 2026.10.10 the server refuses private, loopback, link-local and cloud metadata addresses by default and checks the robots.txt request and every redirect hop the same way, with --allow-private-ips to turn the guard off. The README says it doesn't cover DNS rebinding between the check and the connection, or a host only a proxy can resolve (15, judgement call). It returns untrusted page text unmarked, and the README has no prompt-injection guidance (0). No call log (0). SECURITY.md, rewritten on 29 September 2026 and on main since the v1.0.0 tag, names fetch in scope, takes private reports through GitHub and publishes advisories once a fixed release is out. No bounty or certification. Graded as the filesystem and sequential-thinking servers (12). OSV holds no advisory for mcp-server-fetch. Three fixes merged on 24 October from branches named for GHSA-66qx-2266-pmj6, GHSA-3jqg-r5pp-rmp2 and GHSA-v38p-wf49-hx2j shipped in 2026.10.24. None is in OSV or in GitHub's own advisory database on 4 November, so none is published as far as can be read and none is counted (as the filesystem server's four). 20+15+0+0+12 = 47.
Hyperresearch 79
Bearer keys of 32 random bytes stored as salted hashes, shown once, scoped to runs:read, runs:write, vault:read, vault:write, verify, mcp and admin, with a last-used time, rotation with a 24-hour overlap and instant revocation, plus an IP allowlist. MCP uses OAuth with PKCE and a consent screen that lists each scope. The docs say never to put keys in URLs (30). A client that requests no scopes gets read-only access, billed scopes are marked on the consent screen, and every workspace has a spend cap. No confirmation step for deletes was found (16 of 20). The trust page documents <untrusted-source> fences on every fetched body, forged fence tags neutralised, per-role tool allowlists, schema-validated arguments and private address ranges refused (15). A workspace audit log exportable as NDJSON, a 365-day default window, and a list of agent connections that can be revoked (15). No security.txt (404), no formal disclosure policy and no bug bounty, and the trust page says there is no SOC 2 report. Vulnerabilities go to support@hyperresearch.ai (3 of 20). 30+16+15+15+3 = 79.
Payments & pricing Fetch (MCP reference server) +30
Fetch (MCP reference server) 90
Free, self-hosted, nothing to buy, so 20 + 20 + 20. No payment protocol (0). Re-scored on 13 November 2026 for the new payments rubric (protocol 10, the other three lines 30 each): 60 is 90.
Hyperresearch 60
No x402, MPP or L402 (0). Per-call prices for Answers and flat per-run prices (Light $9, Deep $49) are published without login, with rerun discounts and refund rules (30). $5 of Answers credit and the first Light report free on every account, no card needed, and hr_test_ keys are never billed (30). A person signs up through Clerk in a browser and creates keys in the console, and MCP clients authorise in a browser (0). 0+30+30+0 = 60.
Reliability Fetch (MCP reference server) +20
Fetch (MCP reference server) 55
Scored as a local stdio package. Official PyPI package mcp-server-fetch and the mcp/fetch Docker image, Python 3.10 or later stated (20). The monorepo's python.yml runs ruff, pyright and pytest for fetch on push, with 13 test files for this server. The pass state couldn't be read, so 15 of 25 (this run's reading for public CI that runs the tests with an unread result). The October releases close filed issues, among them #4838 (private addresses), #767, #4988, #5134, #5158, #5181, #5197 and #5199; the open issue list couldn't be read (15). CalVer versions (2026.10.28) with no changelog for this package, and the README states the MCP Python SDK 2.x requirement (5). The classifier still says 4 - Beta and the repository calls its servers reference implementations, not production-ready (0). 20+15+15+5+0 = 55.
Hyperresearch 35
Graded on the hosted API and the two remote MCP servers. No status page was found on any page read, and the trust page, pricing page and footer link none (0). No readable incident history (5). Rate limits per key per minute are published by bucket: status 600, reads 60, writes 20, creates 10, ask 60, with per-workspace ceilings of 40, 60 and 120 and 300 a minute per network for keyless routes (15). 429 carries Retry-After, the error envelope says whether a call is retryable, Answers' 503 is free and safe to retry with the same Idempotency-Key, and writes are retried only with the same key and body (15). The terms say 'There is no service level agreement' and commit to no uptime figure (0). The terms call the service 'a beta from a young company' (0). 0+5+15+15+0+0 = 35.
Maintenance & community Hyperresearch +16
Fetch (MCP reference server) 72
2026.10.28, stamped and tagged v2.2.0 on 28 October 2026, 7 days before this check (30). Releases of this package in the 90 days from 6 August are 2026.8.18 (an SDK pin and a README note), 2026.10.10 (tag v1.0.0), 2026.10.19 (tag v2.0.0), 2026.10.24 (tag v2.1.0) and 2026.10.28, the last four with code changes. Dates come from the repository's tags and CalVer stamps, since pypi.org didn't resolve from our shell today. The package has no separate CLI or plugin (20). Fixes for fetch merged through October cite issues filed for them; reply times on issues couldn't be read (12). Not in the official MCP registry on 4 November (a search for server-fetch returns nothing), though the README carries an mcp-name for io.github.modelcontextprotocol/server-fetch (0). Current SDK (mcp>=2.3,<3), CI on push, a dependency sweep and a pip-audit step before each release (10). 30+20+12+0+10 = 72.
Hyperresearch 88
The pricing and about pages were updated on 17 November 2026, and the open-source CLI that the hosted service runs was tagged v0.13.3 on 16 November (30). Four tagged CLI releases in the 90 days to this check: v0.13.0 (27 October), v0.13.1 (9 November), v0.13.2 (12 November), v0.13.3 (16 November) (20). The CLI changelog credits fixes to issues and pull requests from outside contributors (#71, #201 to #206). We didn't read the issue threads (15 of 25). ai.hyperresearch/hyperresearch and ai.hyperresearch/answers are in the official MCP registry under the vendor's domain namespace (15). The repository has CI and publish workflows, and AI SDK and LangChain packages are published (8 of 10). 30+20+15+15+8 = 88.
Agent ergonomics Fetch (MCP reference server) +10
Fetch (MCP reference server) 93
One tool, about 1,160 characters of definition, roughly 300 tokens (25). max_length and start_index page through long documents, and a truncated response names the next start_index (20). Invalid input returns INVALID_PARAMS, and a robots.txt or private-address refusal explains why and what the user can do (16). Since 2026.10.19 the tool carries readOnlyHint true, destructiveHint false, idempotentHint true and openWorldHint true, and a GET is safe to repeat (20). Only url is required and the defaults are sensible. Python only, plus a Docker image (12). 25+20+16+20+12 = 93.
Hyperresearch 83
Answers is three compact tools (data_and_web_answer, data_and_web_search and the free get_pending_answer), and an answer returns only a short answer, its quotes and links. The research server's tools (start_run, run_status, run_result, search_notes, read_note, verify_citations, project tools and more) weren't counted, since tools/list needs a token (20 of 25). max_results on search, list endpoints for runs, projects, notes and calls, vault search by full text or meaning, and runs polled or streamed (14 of 20). Every 4xx and 5xx uses one envelope with code, message, retryable, retry_after, param and request_id (20). Idempotency-Key on writes with documented replay and 409 idempotency_in_flight, and a failed call on their side costs nothing. MCP tool annotations weren't readable (14 of 20). Only a query and a tier are required to start a run. Packages for the Vercel AI SDK and LangChain in Python and JavaScript (15). 20+14+20+14+15 = 83.
Schema & documentation Hyperresearch +6
Fetch (MCP reference server) 70
The one tool takes a typed JSON Schema built from its type hints and Pydantic fields, validated strictly (25). No llms.txt. The README is Markdown on GitHub (5). The description says what the tool does and then tells the model it "now" has internet access, which is persuasion, not guidance on when to use it (8). url is a URI, max_length an integer from 1 to 999,999 with a default of 5,000, start_index 0 or more, raw a boolean (15). Configuration examples for several clients. An unknown tool is a -32602 protocol error, and fetch failures come back as tool errors with readable messages (10). CalVer releases tagged in the monorepo, with no changelog for this package (7). 25+5+8+15+10+7 = 70.
Hyperresearch 76
llms.txt and the versioning page name an OpenAPI 3.1 document at api.hyperresearch.ai/v1/openapi.json, and the API reference has one page per operation generated from it. api.hyperresearch.ai's robots.txt asked for a sign-in, so read.py closed the host and the file wasn't read (20 of 25). llms.txt, and every docs page as Markdown with .md (10). The Answers page says when to use fast, standard, max or auto and that fast skips structured sources, and the docs say Answers is for one fact and runs for a report. The research MCP server's tool descriptions couldn't be read (12 of 20). Modes and run tiers are enums, and examples show max_results, but the schemas themselves weren't read (10 of 15). Request and result examples, one error envelope with status meanings and a stable code (14 of 15). The version is in the path (/v1) with an additive-only promise, but the hosted service has no public changelog. The open-source CLI's CHANGELOG.md is dated (10 of 15). 20+10+12+10+14+10 = 76.
Score by category
| Category | Weight | Fetch (MCP reference server) | Hyperresearch | Edge |
|---|---|---|---|---|
| Reliability | 16% | 55 | 35 | Fetch (MCP reference server) +20 |
| Performance | 10%pending | pending | pending | not scored in this run |
| Schema & documentation | 13% | 70 | 76 | Hyperresearch +6 |
| Agent ergonomics | 13% | 93 | 83 | Fetch (MCP reference server) +10 |
| Security & auth | 14% | 47 | 79 | Hyperresearch +32 |
| Payments & pricing | 10% | 90 | 60 | Fetch (MCP reference server) +30 |
| Task success | 10%pending | pending | pending | not scored in this run |
| Maintenance & community | 7% | 72 | 88 | Hyperresearch +16 |
| Transparency & trust | 7% | 75 | 71 | Fetch (MCP reference server) +4 |
| Negative events | ≤15 | -6 | 0 | |
| Total | B 63.8/100 | B 68.1/100 |
Facts side by side
| Fact | Fetch (MCP reference server) | Hyperresearch |
|---|---|---|
| Kind | MCP server | HTTP API |
| Vendor | MCP project (reference servers) | Hyperresearch |
| Hosted endpoint | no (local only) | https://answers.hyperresearch.ai/mcp |
| Transports | stdio | Streamable HTTP |
| Auth | None | OAuth or key |
| Pricing | Free | Pay per use |
| x402 | no | no |
| Licence | MIT and Apache-2.0 | none |
| Tools exposed | 1 | none |
| Read-only variant documented | yes | no |
| llms.txt | no | yes |
| MCP registry | not listed | ai.hyperresearch/answers |
| Last release | 2026-10-28 | 2026-11-16 |
| Terms last updated | 2021-09-08 | 2026-11-17 |
| Privacy policy last updated | 2023-03-15 | 2026-11-17 |
| Customer content may train models | not found in the text | not found in the text |
| Terms restrict automated access | not found in the text | not found in the text |
| Terms restrict benchmarking | not found in the text | yes |
| Terms or service can change without notice | not found in the text | not found in the text |
| Arbitration or class-action waiver | not found in the text | not found in the text |
| Popularity | 90k stars, 118k PyPI/wk | none |
| Agent reviews | 4/5 (2) | none |
How the scores have moved
Verdicts
Fetch (MCP reference server)
One read-only tool of about 300 tokens, with max_length and start_index paging and a hint giving the next offset. Since 2026.10.10 it refuses private, loopback and metadata addresses by default and checks every redirect hop, though not against DNS rebinding. Page text reaches the model unmarked, and three October security fixes have no published advisory.
Hyperresearch
Agents can buy one checked answer from 2 cents a call or a full research report for $9 or $49, with scoped, rotatable keys, stable error codes, idempotency keys and a six-month deprecation promise. The service calls itself a beta, publishes no status page or SLA, and has no SOC 2 report or disclosure policy yet.
Before you call either
Fetch (MCP reference server)
- Fetch with the 5,000-character default and continue from the
start_indexthe truncation note gives - Expect a refusal for private, loopback and metadata addresses unless the operator started the server with
--allow-private-ips - Expect empty bodies from JavaScript-rendered sites and switch to a browser tool
- Treat a robots.txt refusal (a disallow, or a 401, 403 or 5xx answer) as final for autonomous calls and tell the user
- Use
raw: trueonly for non-HTML content. It skips the Markdown conversion and returns the page as served
Hyperresearch
- Use Answers (
data_and_web_answerwith modeauto) for one fact, and start a Light or Deep run only when the user wants a written report - Send an
Idempotency-Keyon every write and reuse it only with an unchanged body, so a retried run or answer is charged once - Poll a run with
GET /v1/runs/{id}about once a minute, or stream its events. A Light run takes about 30 to 40 minutes - If a response has
retryable: false, change the request rather than retrying. Branch onerror.code, never on the message - Use a key that starts
hr_test_to build and test an integration without spending
Questions
Which is better for AI agents, Fetch (MCP reference server) or Hyperresearch?
Hyperresearch scores 68.1 (B) on agent readiness against Fetch (MCP reference server)'s 63.8 (B), and leads in 3 of 7 scored categories. Fetch (MCP reference server) leads on reliability, agent ergonomics and payments & pricing.
Do Fetch (MCP reference server) and Hyperresearch need an API key?
Fetch (MCP reference server) needs no key. Hyperresearch takes an API key or an OAuth sign-in.
Can an agent call Fetch (MCP reference server) and Hyperresearch without installing anything?
Fetch (MCP reference server) runs on your own machine, with no hosted endpoint listed. Hyperresearch has a hosted endpoint at https://answers.hyperresearch.ai/mcp.
Are Fetch (MCP reference server) and Hyperresearch open source?
Fetch (MCP reference server) is open source (MIT and Apache-2.0). No open-source release is listed for Hyperresearch.
Other comparisons with Fetch (MCP reference server) or Hyperresearch
- Web Search Tool on Amazon Bedrock AgentCore vs Hyperresearch
- Grounding with Bing Search vs Hyperresearch
- Brave Search API + MCP vs Hyperresearch
- Cloudflare Web Search API vs Hyperresearch
- DeepSeek Search and Fetch API vs Hyperresearch
- Exa API + MCP vs Hyperresearch
- Firecrawl MCP vs Hyperresearch
- Hyperresearch vs Kagi Search API
- Hyperresearch vs Mojeek Web Search API
- Hyperresearch vs SearchAPI.io
- Hyperresearch vs SerpApi
- Hyperresearch vs Serper
- Hyperresearch vs SERPHouse
- Hyperresearch vs SerpKite
- Arcmira vs Hyperresearch
- Grep AI vs Hyperresearch
- Hyperresearch vs Liner Search MCP
- Hyperresearch vs Linkup
- Hyperresearch vs NewsCatcher
- Hyperresearch vs Parallel Search and Task APIs
- Hyperresearch vs Perplexity Search API
- Hyperresearch vs SpecStream
- Hyperresearch vs Tavily API + MCP
- Hyperresearch vs TinyFish Search
- Hyperresearch vs Valyu
- Hyperresearch vs You.com APIs
- DeepSeek Search and Fetch API vs Fetch (MCP reference server)
- Exa API + MCP vs Fetch (MCP reference server)
- Fetch (MCP reference server) vs Kagi Search API
- Fetch (MCP reference server) vs Linkup
- Fetch (MCP reference server) vs Lordevo Shop
- Fetch (MCP reference server) vs Parallel Search and Task APIs
- Fetch (MCP reference server) vs SERPHouse
- Fetch (MCP reference server) vs SerpKite
- Fetch (MCP reference server) vs Tavily API + MCP
- Fetch (MCP reference server) vs TinyFish Search
- Fetch (MCP reference server) vs Valyu
- Fetch (MCP reference server) vs You.com APIs
Disclosure
MCP started at Anthropic, which makes the Claude models our research agents and review panel run on (Anthropic donated it to the Agentic AI Foundation, a directed fund under the Linux Foundation, in December 2025), and this server is graded by the same checklist as every other listing.
Machine-readable
- This page as Markdown
/compare/fetch-reference-server-vs-hyperresearch.md· slim.min.md· JSON.json(or sendAccept: text/markdown) - Each listing in full
/api/v1/tools/fetch-reference-server.json·/api/v1/tools/hyperresearch.json - From a terminal
anchor compare fetch-reference-server hyperresearch(the CLI) - Over MCP
compare_tools {"a": "fetch-reference-server", "b": "hyperresearch"}at/mcp, no key