Exa
$7/ 1,000 searches
Auto Search
- Results
- Up to 10 / search
- Billing
- Usage-based; no subscription
- Contents
- Priced separately
Exa vs Tavily: choose source discovery or research context
Exa
- What it is
- A search API that finds sources by meaning and retrieves their content.
- Best for
- Discovering relevant pages from a question, with domain and date filters.
- Typical output
- Source URLs, page text and query-focused highlights.
Tavily
- What it is
- A web research API for search, page extraction, site crawling and cited reports.
- Best for
- Gathering source context and optional answers for an agent in one search request.
- Typical output
- Ranked sources, relevance scores, content snippets and optional answers.
Need both?
Paste this prompt into your agent.
set up https://glasser.ai/SKILL.md
Show me how to use Exa and Tavily together through Glasser.What can Exa and Tavily do?
| Capability | Exa | Tavily |
|---|---|---|
| Web discovery | ||
| Content retrieval | ||
| Site collection | Not covered here | |
| Research & answers |
Limits & data comparison
| Compare by | Exa | Tavily |
|---|---|---|
| Limits | Up to 25 search requests / second on Developer Direct plan capacity; Glasser Search supports up to 25 results, with 10 in the base charge. | 1,000 requests / minute with production keys Direct Crawl: 100/min; Research creation: 20/min. Search can return up to 20 results. |
| Data fields | URLs, titles, dates and optional page text Use Contents for known URLs; add summaries or highlights when the selected endpoint supports them. | Title, URL, relevance score and matched content Optional raw content and generated answer; Extract reports failed URLs separately. |
| Collection & context controls | Domain and date filters; similar-page discovery Describe the kind of source you need, then fetch the relevant passages. | Search depth, domains, dates and topics Crawl and Map offer path/depth controls; instructions focus collection on a task. |
| Execution & Reliability | Search returns its result in one request Direct Contents supports a live-crawl timeout and per-URL statuses. Handle failed URLs individually. | Search and Extract return in the request; Research is a job Extract has a configurable wait and per-URL failures. Poll Research status; respect Retry-After on 429 responses. |
| Glasser access | Search, similar pages, Contents and Answers Choose a published Glasser operation; provider features can exceed its supported inputs. | Search, Extract, Crawl, Map and Research All five operations are available through Glasser; depth and volume affect the charge. |
Exa vs Tavily pricing
Tavily
$8/ 1,000 basic searches
Pay-as-you-go Search
- Basic search
- 1 credit / request
- Advanced search
- 2 credits / request
- Optional Project plan
- $30/month · 4,000 credits
Task-based cost example
Exa: 1,000 Auto searches with up to 10 results + Tavily: 1,000 basic searches
| Task volume | Direct: entry spend | Glasser: task charges |
|---|---|---|
| Exa: 1,000 Auto searches with up to 10 results | $7.001,000 × $0.007; no subscription | $7.001,000 × $0.007 |
| Tavily: 1,000 basic searches | $8.001,000 × $0.008; no subscription | $8.001,000 × $0.008 |
| Entry spend / task charges | $15.00 | $15.00 |
Inputs and outputs
Exa
- InputA question or description of sources
- OperationSearch by meaning
- OutputRelevant URLs and source snippets
Tavily
- InputA question, domains and search depth
- OperationSearch and collect context
- OutputScored sources and optional answer
Still unsure? Choose based on your task
Find sources that match a concept
Use Exa when you can describe the kind of source you need and want to retrieve selected pages afterward.
Ground an agent answer
Use Tavily when scored snippets and an optional answer fit your research step. Set the search depth to match the task.
Cross-check a developing story
Discover relevant publications with Exa, then use Tavily to gather a second set of sources for your agent to compare.
Why Use Exa and Tavily Through Glasser?
One API Key
Access both providers with one Glasser key.
One Balance
Use one shared balance for both providers.
Pay as You Go
Pay per use, with no Glasser subscription.
Agent-Ready Integration
Connect your agent via Skill, MCP or CLI.
Questions before you choose
Which is better for finding relevant sources?
Exa supports meaning-based discovery and similar-page search. Tavily emphasizes ranked research context and optional answers. Compare the returned sources on your own questions; this page makes no accuracy ranking.
Do I need a separate page-fetch step?
It depends on the output. Exa Contents retrieves known URLs; Tavily Search can include raw content and Tavily Extract handles supplied URLs. Snippets alone may omit details you need to verify an answer.
Can I use Exa and Tavily together in one AI agent?
Yes, for their connected operations. Use Exa to find relevant publications and Tavily to gather research context for a source cross-check. Your agent coordinates the requests using one Glasser key and shared balance.
Can my agent automatically choose between Exa and Tavily?
A tool-using agent can discover endpoints, inspect their inputs and prices, and select an operation for your task. Tell it which output, source filters and request volume you need. Ask for approval before paid calls; Glasser does not make that routing decision for the agent.
How do I control API spending through Glasser?
Search and Inspect are free. Ask your agent to estimate the planned batch, inspect the charge rules and wait for approval before running it. Keep batches small and review final charges; failures can be billable under the endpoint’s rules. The comparison excludes free allowances and does not promise lower unit prices.
Can I use Glasser with Claude Code, Codex and ChatGPT?
Use the Glasser Skill and CLI with local coding agents such as Claude Code or Codex. ChatGPT can connect through the Glasser MCP integration. Follow the setup guide for your client, then inspect the selected operation before making a paid call.