Exa vs Perplexity API for AI agents

As of 15 Oct 2026, Exa gets an AI agent further: L2 · Gets in against L1 · Needs a person.

ExaL2Gets in
  • Find: Docs for agents: llms.txt, an OpenAPI and an MCP server; a docs page for each core task.
  • Access: No key needed, pay per call.
  • Use: 20% of the work.
  • Pay: Pays per call for part of it: Paid $0.001 a call (x402), 15 Oct.
  • Leave: Leaves on its own: Nothing to cancel, made nothing to delete.

In on its own; 20% of the work done. Next: do half of the core work (now 20%)

Gets in describes Exa today: our agent reached search and page-reading with no account at all. It never had to sign up, verify or hold a key, so no checkpoint stopped it on that path. Through the documented keyless hosted MCP server, our agent ran web search and page fetches directly. It also paid per call on Exa's own API, settling x402 charges for search and for fetching page contents, without a dashboard or stored billing details. Reaching an account hit a Cloudflare Turnstile at sign-up. The dashboard turned our agent back at a Vercel checkpoint, so answering, agent runs, websets and monitors stayed out of reach, and cancel and delete never applied.

Gets furtherDriven 15 Oct 2026 · The drive and its evidence
Perplexity APIL1Needs a person
  • Find: Docs for agents: llms.txt, an OpenAPI and an MCP server.
  • Access: Can't sign up.
  • Use: Work not reached.
  • Pay: Payment not reached.
  • Leave: Leaving not reached.

Stopped at sign up: needs a human. Next: let an agent sign up on its own

Gets in on its own fails here: L1 · Needs a person. The docs fetched cleanly headless: llms.txt, markdown pages, and OpenAPI specs. The MCP server's registration also worked headless, but its consent step hit the same challenge. With no key, the API and MCP server answered invalid_api_key, the CLI said no key was configured, and no work got done. No plan was bought, so there was nothing to cancel, but a registered OAuth client has no documented way to delete it.

Driven 6 Oct 2026 · The drive and its evidence

The two summaries: written and checked by our agents from each product's drive. How we write.

Where each one stops.

ExaIn on its own; 20% of the work done

Perplexity APIStopped at sign up: needs a human

Exa and Perplexity API, checkpoint by checkpoint
CheckpointExaPerplexity API
DocsHeadlessHeadless
Sign upNot neededNeeds a human
VerifyNot neededNot reached
KeyNot neededNot reached
First callHeadlessNot reached
First actionNot neededNot reached
PayHeadlessNot reached
CancelNot neededNot reached
DeleteNot neededNot reached

Headless or in a browser.

How our agent passed each checkpoint, and what each product offers an agent.

Checkpoints by route, API, MCP, CLI and SDK
Checkpoints by how our agent passed them, and what each product offers an agent
ExaPerplexity API
Headless (checkpoints)31
In a browser (checkpoints)00
Needs a human (checkpoints)01
Blocked (checkpoints)00
Not tested (checkpoints)07
Work done20% of the corePaid plans only
APIYesYes
OpenAPIYesYes
MCP serverYes (official)Yes (official)
CLINoYes
SDK–Yes
llms.txtYesYes

Bot checks, and where a person was needed.

Walls our agent met
ExaPerplexity API
Bot checkA challenge on the form stopped 2 ways in (Email and code, Sign-in for apps); it got in another wayCloudflare managed challenge (Turnstile 'Verify you are human') stopped it at sign up, with a visible challenge
Wall typeA challenge on the formA check of our network and browser
First wallNoneCloudflare Turnstile on console sign-up
A person helpedNoNo

A challenge on the form: It asks every visitor to prove they are human. It stops every agent on any network, and a signature doesn't help: a person does this step, or the owner opens it for agents. A check of our network and browser: It judged where our agent connects from and its browser, not a person. An agent that works on its own from a data center doesn't get through unless the owner lets signed agents in. How agents in a person's own browser fare isn't measured yet. Where we drive from.

Level and part scores.

Exa and Perplexity API: level and part scores, 0 to 100
ExaPerplexity API
LevelL2 · Gets inL1 · Needs a person
Holds up60 of 100, fairNot measured
Agent-ready setup100 of 100, good100 of 100, good
StandardsMCP spec of 25 Nov 2025 (1 version behind) · MCP discovery (behind) · MCP client registration (current) · OpenAPI 3.1.0 (1 version behind) · A2A 0.3 (1 version behind)MCP client registration (current) · OpenAPI 3.1.0 (1 version behind) · A2A 0.3 (1 version behind)

Docs match is measured in an Audit. What the scores mean.

Pay, cancel and delete.

Paying, cancelling and deleting the account
ExaPerplexity API
PayPays per call for part of itPayment not reached
Free plan or trialFree planFree plan
CancelNot neededNot reached
Delete the accountNot neededNot reached
How each one takes payment
  • Exa: Pays per call for part of it: paid $0.001 a call (x402), 15 Oct.
  • Perplexity API: Payment not reached.

Exa vs Perplexity API: the short answers.

Which can an AI agent use, Exa or Perplexity API?

Exa gets an AI agent further: L2 · Gets in against L1 · Needs a person. With Exa our agent got in on its own and did 20% of the work; with Perplexity API it was stopped at sign up (needs a human). Driven 15 Oct 2026 (Exa) and 6 Oct 2026 (Perplexity API).

Can an AI agent sign up for Exa and Perplexity API on its own?

Exa: not needed. Perplexity API: no, needs a human.

Do Exa and Perplexity API have an MCP server?

Exa: yes (official). Perplexity API: yes (official).

Can an AI agent delete its Exa or Perplexity API account?

Exa: not needed. Perplexity API: not reached.

The drives and the category.

Put your own shortlist side by side.

A Stack check drives the tools you're choosing between and shows which ones your agents can use. $490 for three tools, ready in 24 hours.