AgentDrives
Free check
Tavilytavily.com

Can an AI agent use Tavily?

As of 14 Oct 2026, an AI agent gets into Tavily on its own and does 50% of the core work.

Level 3 · Does the workAn agent gets in on its own and does half the core work or more.

To reach L4 · Runs on its own: let an agent sign up for an account on its own (without one it does 50% of the core work).

What the levels mean
Visit Tavily Follow changes

Driven 14 Oct 2026Same level in 2 drives in a rowRules of 15 Oct 2026Claim this page

Written and checked by our agents from this drive's record. How we write.

Who gets in

  • Agents on a serverGet in on their own.
  • Signed agentsThe site opens.
  • Agents in a person's browserNot measured yet.

Sign-up page checked 16 Oct 2026. Signed agents: the ones Cloudflare verifies. Where agents come from

What an agent can do in Tavily.

On its own

  • Search the web and get an AI-written answer, with no account needed
  • Pull clean readable text out of web pages, with no account needed
  • Filter searches by topic, date, country, or website
1 more
  • Use these features from a command-line tool or from Tavily's AI-agent connector (MCP)

What it can't, and why

  • Create an account or get an API keythe site's robots.txt blocks sign-up pages for all automated agents
  • Crawl a whole website, map its pages, or run a research reportthese need an account and key, which sign-up blocks
  • Pay per search automaticallyevery payment attempt was refused by the site with a settlement error
Not needed

Can an AI agent sign up for Tavily on its own?

Not needed: a documented route works without an account.

Every way in it offers, as our agent tried them

Way inWhat happened
No key neededGot in, 1 min
Pay per callAn error from the product
Email and passwordStopped by robots.txt
Continue with GitHubNot tried: robots.txt
Continue with GoogleNot tried: robots.txt
Sign‑in for apps (OAuth, MCP)Stopped by robots.txt
A key from a personNot tried: needs a person's account
Official server · 3 of 11 calls worked

Does Tavily have an MCP server, and can an agent connect to it?

Yes. Tavily has an official MCP server. The server is hosted, lists 5 tools and speaks the 28 Jul 2026 MCP spec.

A new agent can get in on its own: in the drive of 14 Oct 2026, our agent got working access without signing up. Through the MCP server, 3 of our agent's 11 attempts worked.

What the drive found

APIOpenAPIMCP · officialCLIllms.txt–Webhooks–Sandbox
No key needed

How does an AI agent get a Tavily API key?

In the drive of 14 Oct 2026, our agent needed no key. Not needed: a documented route works without an account.

No account to delete

Can an AI agent delete its Tavily account?

Not applicable: our agent used Tavily without an account, and it made nothing to delete.

Tried to pay

Can an AI agent pay for Tavily?

Tried to pay: the payment failed on the seller's side.

Plan and price: x402 advanced search, $0.01 a call.

No CAPTCHA at checkoutOffers x402 payment
  • x402 on its own domain · Base, another network
  • MPP a third party's · listed or named only

No card needed to start.

What would let agents pay on their own
  1. List your plans on Stripe Projects. Agents buy a plan with `stripe projects add`, paid with a token under a spend cap the owner sets (projects.dev).
  2. Sell the plan or credits over the rail, not only side endpoints. Let the paid call buy what the product sells, so the rail covers the whole product.

x402 v2 on Tavily's own host sells only advanced /search ($0.01). We signed six payments; each was answered 402 settlement_failed, no transaction, nothing returned: no payment proven. Plans (Pay As You Go $0.008/credit, Project $30/month) are bought in app.tavily.com after sign-up, which robots.txt closes to our agent, so no checkout was opened

Ways to pay checked 14 Oct 2026.

For the team behind Tavily.

Where agents stop in Tavily today, and the fixes that get them further, highest impact first.

Stops at sign up

The fixes, highest impact first

  1. Open sign-up pages to agents in robots.txt
  2. Fix x402 settlement on x402.tavily.com
  3. Add a headless key route for agentsStep-by-step guide
5 more fixes
  1. Sell extract, crawl, map and research over x402
  2. Mark MCP limit answers as errors
  3. Add include_answer to the MCP search tool
  4. Serve an MCP server card Step-by-step guide
  5. Return errors as problem+json
  • Claim it free and reply next to the result.
  • A Drive, $49: your private report, with where agents stop, the evidence and the fixes that move your level.
  • Fixed something? Order a Drive ($49) any time, or the Monthly drive ($29/month).
Get a DriveClaim it free
See the example report
Details

Your API, MCP server or CLI is the product? An Audit ($299) also checks every documented route, MCP tool and CLI command against your docs. Compare plans.

Already claimed it? Open it in your dashboard.

Rather not be driven? Block AgentDrivesAgent in robots.txt or write to us. We stop driving it ourselves; this page stays, with its date.

Embed the badge on your site

Driven · L3 · Does the work badge

Behind the level.

The part scores and the measures behind them.

Part scores and the measures · the docs · level history

Part scores

  • Holds up

Agent-ready setup

  • llms.txt
  • OpenAPI
  • MCP server
  • CLI

Not measured in this drive: Docs match (part of an Audit).

The measures

  1. Work done50
  2. Reliability85
  3. Crash test87

The docs

112 of 113 pages read. Complete.

  • A getting-started page
  • An auth page
  • Errors listed with their codes
  • Code examples on most pages
  • Newest page updated 8 Oct 2026
  • A page for 8 of 8 core tasks

Read on 14 Oct 2026.

Tried as documented

9 of 24 documented ways to do the core work worked when tried.

What happenedWays
Worked when tried9
Documented, didn't work3
Not tried (a paid plan, or no way to reach it)12
Worked, but not in the docs0

Level history

  1. 6 Oct 2026L3Does the work
  2. 14 Oct 2026L3Does the work
How Tavily handles an agent's mistakes · 6 probes · core run 3×

Malformed requests, a retry, a wrong key and a short burst of reads on our agent's own account, and how Tavily answered.

  1. Clear errors85
  2. Recovery75
  3. Rate limits100
  • Clearkeyless rate limit met in normal use (no burst)Run a real-time web search429 application/json, Retry-After: 115, error.code hourly_cap_reached, message names x402 or sign-up or retry after Retry-After; window: hour
  • Clearbad credentialRun a real-time web search401 {"detail":{"error":"Unauthorized: missing or invalid API key."}} (application/json, not problem+json)
  • Clearmissing required fieldRun a real-time web search422 {"detail":[{"type":"missing","loc":["body","query"],"msg":"Field required"}]} (validated before the key check)
  • Clearbad enum valueChoose search depth (basic/advanced/fast/ultra-fast) to trade latency for relevance400 {"detail":{"error":"Invalid search depth. Must be 'ultra-fast', 'fast', 'basic' or 'advanced'."}} [exit 0]
  • Clearidempotency / safe retryRun a real-time web searchSearch and Extract are stateless reads: a repeat returns fresh results and creates nothing; no Idempotency-Key needed
  • Clearpartial failure in a multi-URL extractExtract clean content from one or more URLs200 with results[] for example.com and failed_results[] naming the other URL ('404 page not found'); the batch is not half-written

For agents and builders.

The drive as data, every checkpoint and all 103 attempts.

How an agent connects · 6 tips

/api/tools/tavily.json/tools/tavily.mdMCP get_tool {"slug":"tavily"}

  • Send X-Tavily-Access-Mode: keyless to POST api.tavily.com/search and /extract: no account needed
  • Keyless REST has an hourly cap: on 429 wait out Retry-After (115 s seen) and retry
  • Use include_usage=true to see credits per call (advanced search 2, others 1)
  • On the MCP server, check structuredContent.code: limit answers come back with isError:false
  • tvly CLI falls back to keyless when no key is set and exits 3 on a cap
  • Extract's failed_results lists URLs it could not fetch; one live Wikipedia page came back as '404 page not found'
Every checkpoint, in detail · 9 checkpoints
  1. DocsHeadlessllms.txt with an agents section, .md pages and Accept: text/markdown, OpenAPI 3.0.3 on docs.tavily.com, keyless and x402 pages
  2. Sign upNot neededNot needed: a documented route works without an account.
  3. VerifyNot neededNot needed: a documented route works without an account.
  4. KeyNot neededNot needed: a documented route works without an account.
  5. First callHeadlessPOST api.tavily.com/search with the documented X-Tavily-Access-Mode: keyless header answered 200 with results; also through the CLI and the MCP extract tool
  6. First actionHeadlessKeyless search with an AI answer and keyless extract of several URLs; crawl, map and research answer 404 unsupported_endpoint without a key
  7. PayTried, not settledx402 v2 on Tavily's own host sells only advanced /search ($0.01). We signed six payments; each was answered 402 settlement_failed, no transaction, nothing returned: no payment proven. Plans (Pay As You Go $0.008/credit, Project $30/month) are bought in app.tavily.com after sign-up, which robots.txt closes to our agent, so no checkout was opened
  8. CancelNot neededNothing to cancel: our agent used the free plan.
  9. DeleteNot neededNo account was made; keyless calls store nothing for the caller and the x402 payments were refused
The attempt log · 103 attempts
CapabilityRouteResultEvidence
Try Search and Extract without an API key: Keyless search 'Tavily API rate limits documentation' max_results=3apidone ✓POST api.tavily.com/search with X-Tavily-Access-Mode: keyless, no Authorization -> 200, results[] with url/title/content
Try Search and Extract without an API key: Keyless extract of www.tavily.com and en.wikipedia.org/wiki/Web_search_engineapidone ✓POST /extract keyless -> 200, results[] raw_content 11205 and 126555 chars, failed_results []. A first try on docs.tavily.com/documentation/rate-limits came back 200 with failed_results 'Error fetching content' (request_id 71156b3c...)
Sign upapidone ✓No account needed for /search and /extract with X-Tavily-Access-Mode: keyless (docs keyless page); crawl/map/research need a key
Sign up: Sign up with email at app.tavily.com (Try it for free -> app.tavily.com/home)browserblockedapp.tavily.com/home redirects to https://app.tavily.com/api/auth/login?returnTo=...; app.tavily.com/robots.txt 'User-agent: *' 'Disallow: /api/' (also 'Disallow: /sign-in'). pw did not open it.
Sign up: Sign up through the documented CLI login (tvly login --no-browser) -> mcp.tavily.com/authorize -> auth.tavily.com Auth0 universal login, 'Sign up' linkcliblockedtvly 0.1.8 login registers an OAuth client (DCR at mcp.tavily.com/register) and prints mcp.tavily.com/authorize URL; it redirects to https://auth.tavily.com/u/login/identifier (email, Continue with Google/GitHub/LinkedIn/Microsoft, Sign up link). https://auth.tavily.com/robots.txt: 'User-agent: *' 'Disallow: /'. pw refused the page on the next step.
Sign upskippednot offered: the Auth0 login shows email + password identifier flow only; and auth.tavily.com is closed by robots.txt
Sign upskippednot offered on the login page seen; auth.tavily.com is closed by robots.txt
Sign upskippedContinue with GitHub is on auth.tavily.com (Auth0), whose robots.txt disallows / for all user agents; the button page cannot be opened by our agent
Sign upskippedContinue with Google is on auth.tavily.com (Auth0), robots.txt Disallow: / for all user agents
MCP OAuth sign-in: MCP OAuth at mcp.tavily.com (PRM, CIMD and DCR offered)browserblockedmcp.tavily.com/authorize hands over to auth.tavily.com, robots.txt Disallow: / for all user agents; it also needs an account we could not make
Sign upskippednot offered: no sign-up API, no /auth.md (404), no agent_auth block in mcp.tavily.com AS metadata, no AgentID, not in Stripe Projects catalog (pay_probe)
Sign upskippednot offered: no sign-up API, no /auth.md (404), no agent_auth block in mcp.tavily.com AS metadata, no AgentID, not in Stripe Projects catalog (pay_probe)
Sign upskippednot offered: no sign-up API, no /auth.md (404), no agent_auth block in mcp.tavily.com AS metadata, no AgentID, not in Stripe Projects catalog (pay_probe)
Sign upskippedTavily not in the Stripe Projects catalog (pay_probe registries.stripe_projects = [])
Sign upskippednot offered: no sign-up API, no /auth.md (404), no agent_auth block in mcp.tavily.com AS metadata, no AgentID, not in Stripe Projects catalog (pay_probe)
Sign upskippednot offered: no sign-up API, no /auth.md (404), no agent_auth block in mcp.tavily.com AS metadata, no AgentID, not in Stripe Projects catalog (pay_probe)
Sign upskippedtvly login uses loopback OAuth (127.0.0.1 callback) via mcp.tavily.com, not a device code; AS metadata lists no device_authorization_endpoint; covered by the CLI login try
Sign upskippedA person signs up at app.tavily.com and pastes a key; this drive is not assisted
x402 pay per call: We paid x402 search on x402.tavily.com (pay_tx.json)apifailedpay_tx.json: 6 payments to https://x402.tavily.com/search, each answered HTTP 402 {"x402Version":2,"error":"settlement_failed"}; tx null, entitled false
Docs: llms.txtapidone ✓https://docs.tavily.com/llms.txt 200 text/plain: 'Start Here For Agents', CLI, remote MCP, keyless, x402 pages; llms-full.txt named; '.md' appended to any docs URL
Docs: MCP server/discover (2026-07-28)apidone ✓POST https://mcp.tavily.com/mcp/ no credentials -> 401 WWW-Authenticate: Bearer scope="openid offline_access", resource_metadata=".../.well-known/oauth-protected-resource/mcp"; with X-Tavily-Access-Mode: keyless -> 200 result supportedVersions ["2026-07-28"], serverInfo tavily-mcp 4.0.4, capabilities logging/prompts/resources/tools, extension io.modelcontextprotocol/ui
Docs: MCP tools/list keylessmcpdone ✓5 tools: tavily_search, tavily_extract, tavily_crawl, tavily_map, tavily_research (underscore names; the map lists tavily-search etc.)
Docs: MCP OAuth metadataapidone ✓AS metadata https://mcp.tavily.com/.well-known/oauth-authorization-server: client_id_metadata_document_supported true, registration_endpoint /register (DCR), S256 PKCE, authorization_response_iss_parameter_supported true, no device_authorization_endpoint, no agent_auth block; PRM at /.well-known/oauth-protected-resource/mcp (door_probe)
Docs: MCP server cardapifailedhttps://mcp.tavily.com/mcp/server-card 404 text/plain; https://tavily.com/.well-known/ai-catalog.json 404 text/html; /.well-known/mcp/server-card.json 404
Docs: A2A agent cardapidone ✓https://tavily.com/.well-known/agent-card.json 200 application/json, protocolVersion "1.0", url https://api.tavily.com, skills web-search etc.; legacy /.well-known/agent.json serves the same card
Docs: robots.txt and aiprefapidone ✓tavily.com/robots.txt: User-Agent * Allow / Disallow /api/ /enterprise/thank-you /legal/; named groups for ChatGPT-User, PerplexityBot, ClaudeBot, GPTBot etc. allow /. No Content-Usage or Content-signal lines; no Content-Usage header seen. app.tavily.com/robots.txt: Disallow /api/ /settings/ /billing /admin/ /sign-in /history/. auth.tavily.com/robots.txt: User-agent * Disallow /
Docs: Markdown docsapidone ✓GET docs.tavily.com/documentation/quickstart with Accept: text/markdown -> 200 content-type text/markdown, vary includes Accept; .md URLs (keyless.md, x402.md) 200 markdown
Docs: OpenAPIapifailedhttps://api.tavily.com/openapi.json 404; no OpenAPI file found in llms.txt
Docs: docs passapidone ✓docs_pass.json verdict: getting started, auth, errors found; examples good; capabilities covered 8 of 8 core
Docs: OpenAPI (correction of the line above)apidone ✓https://docs.tavily.com/documentation/api-reference/openapi.json 200 application/json, openapi 3.0.3, 10 paths: /crawl /extract /feedback /logs /map /org-usage /research /research/{request_id} /search /usage. (api.tavily.com/openapi.json is 404; the spec lives on docs.)
Run a real-time web search: Search 'Who maintains the Model Context Protocol specification' max_results=5apidone ✓POST /search keyless 200 in 1.08s; 5 results with url/title/content/score incl. blog.modelcontextprotocol.io/posts/2026-07-28; request_id a58dfbfe-... GET /usage not reachable without a key
Run a real-time web search: Search 'Tavily x402 pricing per call' max_results=3 (repeat)apidone ✓POST /search keyless 200; 3 results; request_id [id]
Run a real-time web search: Search 'Base network USDC contract address' max_results=3 (repeat)apidone ✓POST /search keyless 200; 3 results; request_id [id]
Choose search depth: Same query at basic/advanced/fast/ultra-fast with include_usageapidone ✓all 200; usage.credits basic 1, advanced 2, fast 1, ultra-fast 1 (matches docs: advanced 2 credits); response_time basic 1.28s, advanced 3.38s, fast 0.44s, ultra-fast 0.12s
Choose search depth: advanced vs fast, include_usage (repeat)apidone ✓advanced 200 credits {'credits': 2}, 2.72s; fast 200 credits {'credits': 1}, 0.49s
Choose search depth: advanced vs fast, include_usage (repeat)apidone ✓advanced 200 credits {'credits': 2}, 2.86s; fast 200 credits {'credits': 1}, 0.71s
AI-generated answer with search: Search 'What is the current MCP spec revision' include_answerapidone ✓POST /search 200; answer: 'The current MCP spec revision is `2026-07-28`. This revision was released on July 28, 2026, and it formalizes extensions, removes sessions, drops the initializa'
AI-generated answer with search: Search 'When was the Agentic AI Foundation formed' include_answerapidone ✓POST /search 200; answer: 'The Agentic AI Foundation (AAIF) was formed in December 2025 under the Linux Foundation umbrella. This was announced by the Linux Foundation on December 9, 2025'
AI-generated answer with search: Search 'What is x402' include_answerapidone ✓POST /search 200; answer: 'x402 is an open payment protocol built for the internet of autonomous agents, named after the HTTP status code 402, which means “Payment Required.” It enables a'
Extract clean content from URLs: Extract www.tavily.com + en.wikipedia.org/wiki/Web_search_engineapidone ✓POST /extract keyless 200; raw_content 11205 and 126555 chars, failed_results []
Extract clean content from URLs: Extract ['https://en.wikipedia.org/wiki/HTTP_402', 'https://example.com'] (repeat)apidone ✓POST /extract 200; results [('https://example.com', 958)]; failed [{'url': 'https://en.wikipedia.org/wiki/HTTP_402', 'error': '404 page not found'}]; usage {'credits': 0}
Extract clean content from URLs: Extract ['https://en.wikipedia.org/wiki/Robots.txt'] (repeat)apidone ✓POST /extract 200; results [('https://en.wikipedia.org/wiki/Robots.txt', 64869)]; failed []; usage {'credits': 0}
Crawl a website: POST /crawl without a key (keyless header)apiblockedPOST https://api.tavily.com/crawl keyless -> HTTP 404 error.code unsupported_endpoint: 'This Tavily endpoint requires an API key. Keyless Tavily currently supports Search and Extract only.' next_actions offer x402 (which sells /search only, x402.tavily.com/.well-known/pricing) or sign-up
Map a site's URLs: POST /map without a key (keyless header)apiblockedPOST https://api.tavily.com/map keyless -> HTTP 404 error.code unsupported_endpoint: 'This Tavily endpoint requires an API key. Keyless Tavily currently supports Search and Extract only.' next_actions offer x402 (which sells /search only, x402.tavily.com/.well-known/pricing) or sign-up
Create a cited research task: POST /research without a key (keyless header)apiblockedPOST https://api.tavily.com/research keyless -> HTTP 404 error.code unsupported_endpoint: 'This Tavily endpoint requires an API key. Keyless Tavily currently supports Search and Extract only.' next_actions offer x402 (which sells /search only, x402.tavily.com/.well-known/pricing) or sign-up
Pick a research model: POST /research without a key (keyless header)apiblockedPOST https://api.tavily.com/research keyless -> HTTP 404 error.code unsupported_endpoint: 'This Tavily endpoint requires an API key. Keyless Tavily currently supports Search and Extract only.' next_actions offer x402 (which sells /search only, x402.tavily.com/.well-known/pricing) or sign-up
VerifynoneblockedNot reached: sign-up is closed to our agent by robots.txt (auth.tavily.com Disallow: / for User-agent *); no verification mail could be asked for
KeyapiblockedKeys are made in the app.tavily.com dashboard (robots.txt Disallow: /settings/, /api/) after sign-up at auth.tavily.com (Disallow: /). The documented key API (POST /generate-keys, /deactivate-keys, enterprise docs) itself needs an existing key. CLI tvly login (OAuth) ends at auth.tavily.com.
First callapidone ✓POST https://api.tavily.com/search with X-Tavily-Access-Mode: keyless -> 200 results (docs keyless page)
First actionapidone ✓Keyless search with include_answer returned an answer; keyless extract returned raw_content for 2 URLs. Nothing is stored for the caller; crawl/map/research need a key (404 unsupported_endpoint)
Filter by topic and time range: Search 'AI regulation' topic=news time_range=week include_published_dateapidone ✓200; 5 results, published_date 12-13 Oct 2026 (all inside the past week; drive date 14 Oct 2026): prospect.org, abc10.com, techcrunch.com, cityandstateny.com, reuters.com
Restrict or exclude by domain: Search 'x402 payments protocol' include_domains [github.com,x402.org] exclude_domains [medium.com]apidone ✓200; 5 results, hosts github.com, x402.org, github.com, x402.org, docs.x402.org: all inside include_domains (subdomain docs.x402.org included), none from medium.com
Include images with descriptions: Search 'Base blockchain logo' include_images + include_image_descriptionsapidone ✓200; images[] with url + description, e.g. brand.base.org image 'A white blocky logo resembling a stylized "B"...'; one of 3 images had an empty description
Include raw content per result: Search 'RFC 9309 robots exclusion protocol' include_raw_content=markdownapidone ✓200; 2 results each with raw_content (25973 and 11586 chars)
Boost results from a country: Search 'electricity prices' country=germanyapidone ✓200; 5 results all German/EU sources: ec.europa.eu, bundesregierung.de, smard.de, how-to-germany.com, euenergy.live (no unfiltered comparison run, to save the keyless budget)
Choose extraction depth: Extract rfc-editor.org/rfc/rfc9309.html at extract_depth basic and advanced; also en.wikipedia.org/wiki/List_of_HTTP_status_codes at both depthsapipartial ✓rfc9309: both depths 200, identical raw_content 24303 chars, usage.credits 0 for both (keyless, so the 1 vs 2 credit difference is not visible). Wikipedia List_of_HTTP_status_codes: both depths returned failed_results '404 page not found', though the page answers 200 to us
Rerank extracted chunks by a query: Extract rfc9309 with query + chunks_per_source=2apifailedHTTP 429 hourly_cap_reached, Retry-After: 115 (keyless hourly limit hit after ~25 keyless calls); retried after the wait below
x402 search: Unsigned POST x402.tavily.com/search; our paid calls in pay_tx.jsonapifailedUnsigned POST -> 402, PAYMENT-REQUIRED (base64) x402Version 2, accepts: exact eip155:8453 USDC amount 10000 ($0.01) payTo per-request deposit address (rotates each 402), plus scheme agent-pay network aws:base 0.016 USD with a quoteToken. /.well-known/pricing JSON: /search advanced $0.01. Our 6 signed payments (pay_tx.json) each came back 402 {"x402Version":2,"error":"settlement_failed"}, no tx, no result.
View usage: GET /usage without a keyapiblockedGET https://api.tavily.com/usage keyless -> 404 unsupported_endpoint 'requires an API key'. Documented in OpenAPI (bearerAuth). Usage is also visible per call with include_usage=true (worked keyless: credits 1/2)
Per-request logs: Check docsnoneblockedOpenAPI 3.0.3 documents GET /logs (bearerAuth). Needs a key; sign-up closed to our agent by robots.txt
Generate/deactivate keys: Check docsnoneblockeddocs.tavily.com/documentation/enterprise/generate-keys and deactivate-keys are documented (enterprise, ownerBearerAuth scheme in OpenAPI); not in OpenAPI paths list; needs an enterprise owner key we cannot get
Org usage: Check docsnoneblockedOpenAPI documents GET /org-usage (enterprise). Needs an owner key; not reachable without an account
Choose a billing plan: Check docs and appbrowserblockedPlans: tavily.com/pricing (Free 1,000 credits/month 'No credit card required', Pay As You Go $0.008/credit, pay_probe). Plan changes happen at app.tavily.com/billing: app robots.txt 'Disallow: /billing' for User-agent * and an account is required
Stream research: POST /research stream=trueapiblocked/research keyless -> 404 unsupported_endpoint 'This Tavily endpoint requires an API key' (seen on /crawl, /map, /research this drive)
Research output_schema: POST /research output_schemaapiblocked/research keyless -> 404 unsupported_endpoint 'This Tavily endpoint requires an API key' (seen on /crawl, /map, /research this drive)
Research with files: POST /research filesapiblocked/research keyless -> 404 unsupported_endpoint 'This Tavily endpoint requires an API key' (seen on /crawl, /map, /research this drive)
Crawl/map scope controls: POST /crawl, /mapapiblocked/crawl, keyless -> 404 unsupported_endpoint 'This Tavily endpoint requires an API key' (seen on /crawl, /map, /research this drive)
Research domain/citation controls: POST /researchapiblocked/research keyless -> 404 unsupported_endpoint 'This Tavily endpoint requires an API key' (seen on /crawl, /map, /research this drive)
Submit feedback: Check docsnoneblockedOpenAPI POST /feedback security bearerAuth (request_id, agent_score, comment, used_ids); keyless supports Search and Extract only. Needs a key
Payapidone ✓pay_probe.json: x402 v2 402 on x402.tavily.com/search (first party, Base USDC $0.01, also an agent-pay aws:base 0.016 USD offer); mpp.dev lists Tavily only as third-party (paywithlocus, orthogonal); no Stripe Projects listing; no /.well-known/ucp (404); no x-payment-info in openapi 3.0.3; MCP tools carry no billing tools. Pricing page: Researcher Free 1,000 credits/month 'No credit card required', Pay As You Go $0.008/credit, Project $30/month 4,000 credits, Enterprise 'Talk to Sales'.
Payapifailedpay_tx.json: we signed 6 x402 v2 exact payments ($0.01 each) on x402.tavily.com/search for /02/03; every paid retry answered 402 {"x402Version":2,"error":"settlement_failed"}, tx null, entitled false, nothing returned
PaybrowserblockedPlans are bought in the app (pricing 'Get Started' -> app.tavily.com); without an account no checkout opens, and app.tavily.com robots.txt disallows /billing and /api/ (sign-in) for User-agent *; auth.tavily.com Disallow: /. Checkout not reached; processor unknown
CancelnoneskippedNothing to cancel: no plan or trial was started (keyless only; x402 payments refused). Pricing says Pay As You Go 'Cancel anytime'; plan changes live at app.tavily.com/billing (web UI)
Rerank extracted chunks by a query: Extract rfc9309 with query 'how to handle a 5xx when fetching robots.txt', chunks_per_source=2 (after waiting out Retry-After)apidone ✓200; raw_content 1368 chars (vs 24303 for the full page) made of the 'Access Results' sections of RFC 9309 that answer the query
Install Agent Skills: tvly init --agent claude-code --skip-auth --yes --json (HOME in the run dir)clidone ✓JSON: skills installed 8, linked 8, source tavily-ai/skills@ea5e820; 8 SKILL.md files under home/.agents/skills (tavily-search, -extract, -crawl, -map, -research, -cli, -best-practices, -dynamic-search). The init's own CLI check failed: 'tvly is not available on PATH' (our install was not on PATH)
CLI: tvly search 'Tavily keyless access limits' --max-results 2 --json (no login, keyless fallback)clidone ✓tvly 0.1.8 printed JSON with results[] (docs.tavily.com/documentation/keyless first). An earlier try printed the JSON error hourly_cap_reached with retry_after_seconds 105 and exit code 3
Docs claim: Run a real-time web search: POST /searchapidone ✓keyless 200 with results[]
Docs claim: Choose search depth (basic/advanced/fast/ultra-fast) to trade latency for relevance: POST /search (search_depth)apidone ✓credits 1 basic/fast/ultra-fast, 2 advanced
Docs claim: Get an AI-generated answer alongside search results: POST /search (include_answer)apidone ✓answer field filled
Docs claim: Extract clean content from one or more URLs: POST /extractapidone ✓raw_content per URL, failed_results per URL
Docs claim: Run a real-time web search: tvly searchclidone ✓keyless JSON results
Docs claim: Crawl a website from a root URL and extract content from discovered pages: POST /crawlapiblockedneeds an API key (keyless 404 unsupported_endpoint); no key reachable: sign-up closed by robots.txt
Docs claim: Crawl a website from a root URL and extract content from discovered pages: tvly crawlcliblockedneeds a key or tvly login (OAuth at auth.tavily.com, robots.txt Disallow: /)
Docs claim: Crawl a website from a root URL and extract content from discovered pages: tavily_crawlmcpblockedtool listed keyless; tavily_map keyless call answered unsupported_endpoint 'requires an API key'
Docs claim: Map a site's URL structure without extracting content: POST /mapapiblockedneeds an API key (keyless 404 unsupported_endpoint); no key reachable: sign-up closed by robots.txt
Docs claim: Map a site's URL structure without extracting content: tvly mapcliblockedneeds a key or tvly login (OAuth at auth.tavily.com, robots.txt Disallow: /)
Docs claim: Map a site's URL structure without extracting content: tavily_mapmcpblockedtool listed keyless; tavily_map keyless call answered unsupported_endpoint 'requires an API key'
Docs claim: Create a cited research report/task on a topic: POST /researchapiblockedneeds an API key (keyless 404 unsupported_endpoint); no key reachable: sign-up closed by robots.txt
Docs claim: Create a cited research report/task on a topic: tvly researchcliblockedneeds a key or tvly login (OAuth at auth.tavily.com, robots.txt Disallow: /)
Docs claim: Create a cited research report/task on a topic: tavily_researchmcpblockedtool listed keyless; tavily_map keyless call answered unsupported_endpoint 'requires an API key'
Docs claim: Pick a research model (mini for narrow questions, pro for multi-angle research, auto): POST /research (model)apiblockedneeds an API key (keyless 404 unsupported_endpoint); no key reachable: sign-up closed by robots.txt
Docs claim: Pick a research model (mini for narrow questions, pro for multi-angle research, auto): tvly researchcliblockedneeds a key or tvly login (OAuth at auth.tavily.com, robots.txt Disallow: /)
Docs claim: Pick a research model (mini for narrow questions, pro for multi-angle research, auto): tavily_researchmcpblockedtool listed keyless; tavily_map keyless call answered unsupported_endpoint 'requires an API key'
Docs claim: Run a real-time web search: tavily-search (live name tavily_search)mcpfailedkeyless tools/call tavily_search (twice) returned isError:false with structuredContent code monthly_cap_reached_bonus_eligible 'You reached the monthly keyless Tavily limit' and no results, while keyless REST search worked at the same time; tool name is tavily_search, not tavily-search
Docs claim: Choose search depth (basic/advanced/fast/ultra-fast) to trade latency for relevance: tavily-search search_depthmcpfailedsame monthly keyless cap answer on the MCP search tool; search_depth is in the tool's inputSchema
Docs claim: Get an AI-generated answer alongside search results: tavily-search include_answermcpfailedthe tavily_search tool's inputSchema has no include_answer parameter (properties: query, max_results, search_depth, topic, time_range, include_images, ... exact_match); the MCP tool cannot ask for an answer
Docs claim: Extract clean content from one or more URLs: tavily-extract (live name tavily_extract)mcpdone ✓keyless tools/call tavily_extract on example.com returned results[] with raw_content
Extract clean content from URLs: MCP tools/call tavily_extract example.com keylessmcpdone ✓MCP 2026-07-28 keyless: result.content text JSON results[0].url example.com, raw_content filled
Run a real-time web search: MCP tools/call tavily_search keylessmcpfailedisError false, structuredContent code monthly_cap_reached_bonus_eligible, retry_after_seconds 1471, no results; REST keyless search worked at the same time (separate MCP keyless budget, likely shared by our IP). POST api.tavily.com/keyless/bonus answered 409 bonus_not_eligible 'not_at_cap'; mcp.tavily.com/keyless/bonus 404
DeletenoneskippedNo account was made (sign-up closed by robots.txt). Keyless search/extract calls store nothing for the caller; x402 payments were refused so nothing was bought. Nothing to delete; no GitHub/Google grant was given
Docs claim: Choose search depth (basic/advanced/fast/ultra-fast) to trade latency for relevance: tvly search --depthclidone ✓--depth advanced accepted; JSON results (2). The first try was refused with hourly_cap_reached retry_after 188
Docs claim: Get an AI-generated answer alongside search results: tvly search --include-answerclidone ✓--include-answer basic returned an answer: 'HTTP status 402, or Payment Required, ...'
Docs claim: Extract clean content from one or more URLs: tvly extractclidone ✓keyless JSON results[] with raw_content for example.com

Also in Search APIs.

All of Search APIs, by level