Skip to the page
Get started free

Exa Research API alternative in 2026

Rows with a source per value.

Exa replaced its /research endpoint with /search and type: "deep-reasoning" in April 2026, at $15 per 1,000 requests. Since Oct 19 that type is legacy, and Exa sends list building to deep or to Exa Agent, billed per run.

SpecStream is the live web data engine for teams and agents. Deep Search takes a question in plain words and returns every match as structured rows, with the source URL and read time on every value, at 5¢, 15¢ or $1 a delivered row.

Describe your data

New accounts get $10 of credit, no card. Reading and pricing a request cost nothing.

SpecStream and Exa's research options side by side

SpecStreamExa
Where /research wentDeep Search, on the same key as the Search API and the Extract API, through the REST API, MCP, Python and TypeScript SDKs and a CLI/search with type: "deep-reasoning", a legacy type since Oct 19, 2026. Exa recommends deep, or Exa Agent for long-running research, list building and multi-hop enrichment
What you sendA request in plain words. SpecStream reads it back as the conditions every row must meet and the facts to bring back, and you can edit each one before it runsA query and an outputSchema. On deep, the schema holds at most 10 properties in total, counted across every level, and 2 levels of nesting
A wide tableName every fact the row needs. The row rate follows how much each row needs: 5¢, 15¢ or $1More than 10 fields means several deep requests with focused schemas, or Exa Agent
The source of each valueSource URL and read time on every value. Each condition shows the page that states it, or that no page states it yetoutput.grounding with field-level citations and a confidence
Price5¢, 15¢ or $1 a delivered row, plus 5¢ for a person's email and 15¢ for a phone$15 per 1,000 requests on deep-reasoning and $12 on deep. Exa Agent: $0.012 to $1.00 a run at a fixed effort, or metered up to a $5 cap on auto
The price before the runThe most the run can cost, quoted before it starts. Nothing runs until you approve it, and reading and pricing a request cost nothingA price per request by type. Metered Agent runs take budget.maxCostDollars from $1 to $100
How long it runsUntil it reaches the count you asked for, up to four hours on the largest listsdeep 4 to 15 seconds and deep-reasoning 12 to 40 seconds. Agent Ultra takes a time budget of 5 minutes to 3 hours
What an empty result costsNothing. A run that finds nothing costs nothing, and duplicates and rows a page contradicts are not chargedEach API call is a request, priced per 1,000 requests
Keeping it currentA saved request reruns hourly to yearly, each row labeled changed, unchanged or no longer matching. An unchanged row costs 1¢Monitors run scheduled searches at $2 per 1,000 requests
Where the rows goSheets, Excel, Notion, Airtable, HubSpot, Salesforce, Pipedrive, Attio, Slack, Teams, webhooks, Zapier, Make, n8n, the API, SDKs and MCPJSON in output.content from /search, or in output.structured from Exa Agent

Why /research users move to SpecStream

What happened to /research

Exa's April 1, 2026 notice replaced /research with /search and type: "deep-reasoning". On Oct 19, 2026 Exa made deep-reasoning a legacy type: it keeps its $15 per 1,000 price, and Exa points new requests to deep at $12, or to Exa Agent for long-running research, list building and multi-hop enrichment. Code that called one research task now picks between a search type and an agent priced per run.

A table as wide as the job

A deep request holds at most 10 properties in its outputSchema, nested objects and array items included, so a list of companies with 4 facts each already uses 5. Past 10, Exa suggests several requests or Exa Agent. On SpecStream you name the facts, and the quote prices the row by how much it needs: 5¢, 15¢ or $1.

A source on every value

Exa returns output.grounding with citations and a confidence per field, and its guide asks you to verify important claims against their sources. On SpecStream every value has its page: the URL it was read on and the time it was read. Each condition of your request shows the page that states it, or that no page states it yet, and get_find_decisions returns the answer, the reason, the value, the source and the time.

Priced per row, before the run

A deep-reasoning call costs $0.015 whatever its output holds, and an Agent run bills compute and searches as it goes. SpecStream prices the delivered row and shows the most the run can cost before it starts. Duplicates, rows you already own and rows a page contradicts are not charged, and what the run does not use goes back to your balance. See Deep Search and the pricing page.

Other Exa Research API alternatives

Exa Agent

Exa's own answer for list building and multi-hop enrichment: $0.012 to $1.00 a run at a fixed effort, or metered at $0.10 per compute unit and $0.005 per search up to a $5 cap on auto, plus $0.02 per email and $0.07 per phone found. SpecStream prices the delivered row instead, with the maximum approved before the run.

Parallel Task API

Research runs priced per 1,000 by processor, from $5 on lite to $2,400 on ultra8x, one charge per run whatever the output fields. Every output carries citations, reasoning and confidence. On SpecStream one request returns every match as its own row, each priced and checked.

Tavily Research

A research endpoint with a JSON Schema output, billed between 4 and 110 credits a request on mini and 15 to 250 on pro, at $0.008 a credit pay as you go. SpecStream names the most a run can cost before it starts, and the charge never goes past it.

Exa Websets

Exa's list-building API, billed in Websets credits kept apart from Search API credits, with the Free plan stopping a webset at 25 verified results. The Exa Websets comparison sets it beside SpecStream row by row.

How SpecStream scores

The same numbers as the benchmarks page, from the official tasks and scorers.

SpecStreamParallelExaTavilyPerplexity
Web researchWideSearch, GPT-5.6 Luna60.945.553.047.947.0
Factual accuracySimpleQA Verified, GPT-5.6 Luna9694919494
Factual accuracySimpleQA Verified, GPT-5.6 Sol9997919295

How to move from /research

1. Write what your research task asked for as one request in plain words: the records you want, the conditions each must meet, and each field of your outputSchema as a fact to bring back.

2. Connect your agent to https://api.specstream.ai/mcp, or create a key under Agents & API in the workspace. New accounts get $10 of credit, no card. At 5¢ a row, that is up to 200 rows.

3. compile reads the request back as editable parts and quote_capacity_work prices it. Show the reading and the maximum to the person who pays, then authorize_capacity_work starts the run under that maximum. list_find_rows reads the rows with their sources.

Or send the request from code with no account. This call spends nothing: the 402 reply shows how SpecStream read the request, the columns and the most it can cost.

curl -i -X POST https://api.specstream.ai/v1/agent/lists \
  -H "Content-Type: application/json" \
  -d '{"request": "Climate software startups in Europe that raised a seed round in 2026, with the round size and the lead investor", "count": 50}'

Pay by card with MPP or in USDC with x402, send the same request again, and read the rows at the status_url. The agent payments guide has each step. For plain search calls, the Exa migration guide moves /search with one line.

Describe your data

Frequently asked questions

Something we didn't answer? Talk to a human

Is the Exa Research API deprecated?
Yes. Exa's April 1, 2026 notice replaced the /research endpoint with /search and type: "deep-reasoning", and on Oct 19, 2026 Exa made deep-reasoning a legacy type that keeps its price of $15 per 1,000 requests.
What replaces Exa's /research endpoint?
Exa points new requests to type: "deep" on /search, at $12 per 1,000 requests, or to Exa Agent for long-running research, list building and multi-hop enrichment, at $0.012 to $1.00 a run at a fixed effort. SpecStream's Deep Search returns the research as structured rows at 5¢, 15¢ or $1 a row.
How many fields can an Exa deep search return?
10 properties in total in its outputSchema, counted across every level, with at most 2 levels of nesting. For more, Exa suggests several deep requests with focused schemas, or Exa Agent.
How much does a research row cost on SpecStream?
5¢, 15¢ or $1 a delivered row, by how much the row needs, plus 5¢ for a person's email and 15¢ for a phone. 200 rows at 15¢ cost at most $30, and the quote shows the maximum before the run starts.
Does SpecStream give a source for each value?
Yes. Every value carries the URL of the page it was read on and the time it was read, and get_find_decisions returns each condition's answer with its reason, value, source and time.
Does SpecStream answer Exa's search call?
Yes. Set the base address to https://api.specstream.ai/exa, and POST /exa/search answers in Exa's JSON for $1 per 1,000 searches of up to 10 results, or $5 on a deep type.

Sources

SpecStream wrote this page. Each rival figure comes from that company's own page.

  1. Exa changelogexa.ai/docs/changelog.md, read Nov 12, 2026
  2. Exa pricingexa.ai/docs/admin/pricing.md, read Nov 12, 2026
  3. Exa Deep Searchexa.ai/docs/search/deep-search.md, read Nov 12, 2026
  4. Exa Agent quickstartexa.ai/docs/agent/quickstart.md, read Nov 12, 2026
  5. Exa Agent best practicesexa.ai/docs/agent/best-practices.md, read Nov 12, 2026
  6. Exa Websets overviewexa.ai/docs/reference/websets-overview, read Nov 12, 2026
  7. Parallel pricingdocs.parallel.ai/getting-started/pricing.md, read Nov 12, 2026
  8. Parallel Task quickstartdocs.parallel.ai/task-api/task-quickstart, read Nov 12, 2026
  9. Tavily creditsdocs.tavily.com/documentation/api-credits, read Nov 12, 2026
  10. SpecStream pricingspecstream.ai/pricing, read Nov 12, 2026
  11. SpecStream for agentsspecstream.ai/agents.md, read Nov 12, 2026