Head to head · Research search · October 2026 research run
Hyperresearch vs Liner Search MCP
Hyperresearch scores 68.1 (B) on agent readiness against Liner Search MCP's 44.7 (E), and leads in 6 of 7 scored categories. Liner Search MCP leads on reliability. Both do research search.
Best web search APIs for AI agents · All 353 search comparisons
Which one, for what
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 45
- Agent ergonomics, 83 against 65
- Security & auth, 79 against 35
- Payments & pricing, 60 against 30
- Maintenance & community, 88 against 45
- Transparency & trust, 71 against 34
Also in its favour
- Free to start without a card
Watch for
No status page, and the terms say there is no SLA and no uptime figure
Best for Agents that need cheap cited search across the web and scholarly sources, with optional deep research reports.
Ahead on
- Reliability, 50 against 35
Watch for
No free tier found
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 +44
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.
Liner Search MCP 35
OAuth 2.1 or an API key; scopes not published (20 of 30). All tools are reads (10 of 20). Results carry web pages and documents with no injection guidance (0). No audit log found (0). No security.txt (404) or security programme found (5 of 20). 20 + 10 + 0 + 0 + 5 = 35.
Maintenance & community Hyperresearch +43
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.
Liner Search MCP 45
Registry entry com.liner/search last updated 11 September 2026 (20). No dated changelog found (0). Support page linked from the docs (10 of 15). Official registry entry under its own namespace (15). No package to read (0). 20 + 0 + 10 + 15 + 0 = 45. Registry presence read from Anchor Terminal's own daily import of the official MCP registry (/api/v1/registry.json, 6 November 2026), not the registry's API.
Transparency & trust Hyperresearch +37
Hyperresearch 71
The hosted service is closed with terms from Thalient Labs Inc. (version 1.7), and the CLI that runs the same pipeline is MIT (20 of 30). The trust page states retention per record type (content kept until deleted, verification inputs 7 days, billing at least 7 years, audit log 365 days by default), names the known gaps (the run index and questions are kept indefinitely, no automatic expiry), says customer content isn't used for training, and says it matches the privacy policy. No DPA on the shelf (26 of 30). A written deprecation policy: at least six months' notice, Deprecation and Sunset headers and a list of deprecations, plus 30 days' notice by email before a shutdown (20). Subprocessors are listed with what each receives and their countries, and the trust page says there is no data residency guarantee on Cloudflare's network (18 of 20). 20+26+20+18 = 84.
Liner Search MCP 34
Terms and privacy linked from the docs but not read in this batch (5 of 30). Data handling not read (5 of 30). No deprecation policy found for the MCP server; the docs note older endpoints priced per version (5 of 20). Subprocessors not read (5 of 20). 5 + 5 + 5 + 5 = 20.
Schema & documentation Hyperresearch +31
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.
Liner Search MCP 45
Tool schemas not read without sign-in (15 of 25). llms.txt (10). A table says which tool fits which need, and that the domain tools return the same shape as search_web (12 of 20). Inputs not verifiable (3 of 15). Setup examples, no error responses (5 of 15). Registry version 1.0.0 since 11 September; no changelog found (0). 15 + 10 + 12 + 3 + 5 + 0 = 45.
Payments & pricing Hyperresearch +30
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.
Liner Search MCP 30
No machine payment protocol (0). Per-call prices public (20). No free tier found (0). A person signs up in a browser for OAuth or an API key (0). 0 + 20 + 0 + 0 = 20. Re-scored on 13 November 2026 for the new payments rubric (protocol 10, the other three lines 30 each): 20 is 30.
Agent ergonomics Hyperresearch +18
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.
Liner Search MCP 65
Nine tools (25). No paging or size controls documented (5 of 20). 429 codes documented for the API (8 of 20). All tools read; annotations not read (12 of 20). One URL with OAuth, few required inputs and an OpenAI-compatible API (15). 25 + 5 + 8 + 12 + 15 = 65.
Score by category
| Category | Weight | Hyperresearch | Liner Search MCP | Edge |
|---|---|---|---|---|
| Reliability | 16% | 35 | 50 | Liner Search MCP +15 |
| Performance | 10%pending | pending | pending | not scored in this run |
| Schema & documentation | 13% | 76 | 45 | Hyperresearch +31 |
| Agent ergonomics | 13% | 83 | 65 | Hyperresearch +18 |
| Security & auth | 14% | 79 | 35 | Hyperresearch +44 |
| Payments & pricing | 10% | 60 | 30 | Hyperresearch +30 |
| Task success | 10%pending | pending | pending | not scored in this run |
| Maintenance & community | 7% | 88 | 45 | Hyperresearch +43 |
| Transparency & trust | 7% | 71 | 34 | Hyperresearch +37 |
| Negative events | ≤15 | 0 | 0 | |
| Total | B 68.1/100 | E 44.7/100 |
Facts side by side
| Fact | Hyperresearch | Liner Search MCP |
|---|---|---|
| Kind | HTTP API | MCP server |
| Vendor | Hyperresearch | Liner |
| Hosted endpoint | https://answers.hyperresearch.ai/mcp | https://platform.liner.com/api/v1/mcp |
| Transports | Streamable HTTP | Streamable HTTP |
| Auth | OAuth or key | OAuth |
| Pricing | Pay per use | Paid |
| Price for research search | $9 per transaction | not published |
| x402 | no | no |
| Tools exposed | none | 9 |
| Read-only variant documented | no | no |
| llms.txt | yes | yes |
| MCP registry | ai.hyperresearch/answers | com.liner/search |
| Last release | 2026-11-16 | none |
| Terms last updated | 2026-11-17 | no document linked |
| Privacy policy last updated | 2026-11-17 | no document linked |
| Customer content may train models | not found in the text | not read yet |
| Terms restrict automated access | not found in the text | not read yet |
| Terms restrict benchmarking | yes | not read yet |
| Terms or service can change without notice | not found in the text | not read yet |
| Arbitration or class-action waiver | not found in the text | not read yet |
How the scores have moved
Hyperresearch
No score change since it was listed.
Verdicts
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.
Liner Search MCP
Liner's MCP server puts cheap per-call web, scholarly and vertical search and cited research agents behind one OAuth URL, at $1 per 1,000 searches. The docs say which tool fits which question, but there's no free tier, no changelog, no injection guidance and no status component for the API.
Before you call either
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
Liner Search MCP
- Use search_scholar for papers and the domain tools for legal, finance, medical or business questions
- Deep research is slower and costs more; use quick_answer_agent for short lookups
- Treat search results and research reports as untrusted content
- Keep under 2 requests a second for scholar search and the agents
Questions
Which is better for AI agents, Hyperresearch or Liner Search MCP?
Hyperresearch scores 68.1 (B) on agent readiness against Liner Search MCP's 44.7 (E), and leads in 6 of 7 scored categories. Liner Search MCP leads on reliability.
Do Hyperresearch and Liner Search MCP need an API key?
Hyperresearch takes an API key or an OAuth sign-in. Liner Search MCP uses an OAuth sign-in.
Can an agent call Hyperresearch and Liner Search MCP without installing anything?
Yes. Hyperresearch has a hosted endpoint at https://answers.hyperresearch.ai/mcp and Liner Search MCP at https://platform.liner.com/api/v1/mcp.
Other comparisons with Hyperresearch or Liner Search MCP
- 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
- Arcmira vs Liner Search MCP
- Exa API + MCP vs Liner Search MCP
- Grep AI vs Hyperresearch
- Grep AI 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
- Liner Search MCP vs Linkup
- Liner Search MCP vs NewsCatcher
- Liner Search MCP vs Parallel Search and Task APIs
- Liner Search MCP vs Perplexity Search API
- Liner Search MCP vs SpecStream
- Liner Search MCP vs Tavily API + MCP
- Liner Search MCP vs TinyFish Search
- Liner Search MCP vs Valyu
- Liner Search MCP vs You.com APIs
- Fetch (MCP reference server) vs Hyperresearch
- Liner Search MCP vs SearchAPI.io
- Liner Search MCP vs SerpApi
- Liner Search MCP vs Serper
- Liner Search MCP vs SERPHouse
- Liner Search MCP vs SerpKite
- Grounding with Bing Search vs Liner Search MCP
Machine-readable
- This page as Markdown
/compare/hyperresearch-vs-liner-mcp.md· slim.min.md· JSON.json(or sendAccept: text/markdown) - Each listing in full
/api/v1/tools/hyperresearch.json·/api/v1/tools/liner-mcp.json - From a terminal
anchor compare hyperresearch liner-mcp(the CLI) - Over MCP
compare_tools {"a": "hyperresearch", "b": "liner-mcp"}at/mcp, no key