# Exa alternative: Exa-compatible search for less

> Keirolabs serves Exa's /search and /contents API, so switching is a base-URL change. What is supported, what isn't, how exa-py's defaults affect your bill, and the other alternatives worth testing.

Canonical: https://keirolabs.cloud/exa-alternative · Updated: 2026-11-20

> **Quick answer.** The best drop-in **Exa alternative** is **Keirolabs**. It serves an **Exa-compatible `/search` and `/contents` API**, so you keep `exa-py` and set `base_url` to Keirolabs.
>
> **Keirolabs search costs $0.44 to $1.33 per 1,000** on a plan against **Exa's $4 to $7**, and search with page text costs $1.33 to $4.00 against about $12 on Exa: **up to 16 times cheaper**. The same key also gives you Keirolabs lite at **$0.03 to $0.10 per 1,000**, Google SERP results and cited answers, and Keirolabs deep scores **96.5% on SimpleQA** against Exa's 95.47%. Signup gives **1,000 free credits with no card**. Exa's deep search types, find-similar, highlights, domain and date filters, categories and structured output are not supported yet; keep those on Exa. Other alternatives: **Parallel** for speed, **Tavily** for LangChain, **Brave** for an independent index.

## Switch in one line

`exa-py` takes a base URL. Point it at Keirolabs and pass a Keirolabs key:

```python
from exa_py import Exa

exa = Exa(
    api_key="keiro_...",  # your Keirolabs key
    base_url="https://api.keirolabs.cloud/api/compat/exa/v1",
)

# contents=False: results only, the cheapest path. exa-py requests page text by default.
results = exa.search("best vector database for RAG", num_results=10, type="keyword", contents=False)
for r in results.results:
    print(r.title, r.url)
```

The SDK sends Exa's `x-api-key` header, which Keirolabs accepts. Over HTTP:

```bash
curl https://api.keirolabs.cloud/api/compat/exa/v1/search \
  -H "x-api-key: $KEIRO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"query": "best vector database for RAG", "numResults": 5}'
```

Responses keep Exa's shape: `results` with `id`, `url`, `title`, `publishedDate`, `author`, `score`, `text`, `summary`, `highlights` and `embedding`, plus `requestId` and `cost`. Keirolabs adds `creditsCharged` and `creditsRemaining`.

## What is supported

| Exa endpoint or option | Keirolabs | Notes |
| :--- | :--- | :--- |
| `POST /search`, `type` keyword, fast or instant, no contents | Yes | Served by Keirolabs fast search. Up to 35 results |
| `POST /search` with `contents` (text or summary), or `type` neural or auto | Yes | Served by Keirolabs search plus content, which returns each page's text. Up to 5 results |
| `includeEmbeddings` | Yes | On content searches |
| `POST /contents` with `urls` | Yes, one URL per call | `livecrawl` maps to a heavier fetch. Exa result `ids` cannot be resolved; pass URLs |
| `type` deep, deep-lite, deep-reasoning | No | Returns `501 not_supported` |
| `contents.highlights` | Not yet | Field is present but not populated |
| `includeDomains`, `excludeDomains`, published-date filters, `category` | Not yet | Ignored |
| `outputSchema` (structured output) | Not yet | |
| `findSimilar`, `/answer` | Not yet | |

The live list is at [api.keirolabs.cloud/api/compat/status](https://api.keirolabs.cloud/api/compat/status).

**Heads-up on cost routing:** `exa-py` asks for page text on every search unless you pass `contents=False`. Requests with text or a summary, or with `type="auto"` or `"neural"`, go to the content path: $1.33 to $4.00 per 1,000 and at most 5 results. For plain ranked results, pass `contents=False` and `type="keyword"` to stay on the fast path at $0.44 to $1.33 and up to 35 results.

## Price: Exa vs Keirolabs

Exa's pricing ([exa.ai/pricing](https://exa.ai/pricing), November 20, 2026): lite and instant search $4 per 1,000, auto and fast $7, deep-lite and deep $12, deep-reasoning $15. Each result past 10 adds $1 per 1,000. Contents cost $1 per 1,000 pages for each type (text, highlights, summary).

| Per 1,000 requests | Exa | Keirolabs Essential | Keirolabs Pro | Keirolabs Startup |
| :--- | :--- | :--- | :--- | :--- |
| Search, results only | $4 (lite) to $7 (auto) | $1.33 | $0.74 | $0.44 |
| Search + page text, 5 results | $12 (auto + 5 text pages) | $4.00 | $2.22 | $1.33 |
| Contents, per URL | $1 per content type | $2.67 | $1.48 | $0.89 |

**Contents alone are cheaper on Exa** when you need a single content type: $1 per 1,000 pages against our $0.89 to $2.67. Our advantage is on search, and on search with text in one call.

### What a month costs

| Searches a month (results only) | Exa auto | Exa lite | Keirolabs |
| :--- | :--- | :--- | :--- |
| 2,500 | $7.50 (after the $10 monthly credit) | free ($10 monthly credit) | free (1,000 credits at signup, once), then $29.99 |
| 20,000 | $140 | $80 | $29.99 (Essential covers about 22,500) |
| 200,000 | $1,400 | $800 | $99.99 (Startup covers about 225,000) |

For plain ranked results without Exa's response shape, Keirolabs' native `/search/lite` costs **$0.03 to $0.10 per 1,000** on a plan.

## When Exa is still the better choice

Exa is a different kind of search engine, and some of what it does has no equivalent here:

- **Find-similar:** give Exa a URL and get pages like it. Useful for competitor discovery, lead lists and research.
- **Neural ranking on fuzzy queries:** Exa embeds the query and ranks by meaning. Keirolabs serves `neural` requests with keyword-and-content search, which is not the same thing.
- **Categories and date filters:** research papers, companies, news, people; published-date windows.
- **Deep search and structured output:** Exa's deep types and `outputSchema` return synthesized, typed results.

If your product depends on those, keep them on Exa and move plain search and page fetches to a cheaper API.

## Other Exa alternatives

| Alternative | Best for | Search price per 1,000 | Free to start |
| :--- | :--- | :--- | :--- |
| **Keirolabs** (Exa-compatible) | Same API for search and contents, lower price | $0.44 to $4.00; lite $0.03 to $0.10 | 1,000 credits, no card |
| **Parallel** | Lowest latency, deep research, list-building | $1 (turbo) to $5 | up to $80 + $5 a month, card required |
| **Tavily** | LangChain-native agent search | $8 basic, $16 advanced (pay as you go) | 1,000 credits a month, no card |
| **Perplexity Search API** | Perplexity's index | $1 (fast) to $5 | none listed |
| **Brave Search API** | Independent web index | $5 | $5 a month, card required |

Prices from each vendor's pricing page, November 20, 2026. Full four-way comparison: [Parallel vs Exa vs Tavily vs Perplexity](https://keirolabs.cloud/parallel-vs-exa-vs-tavily).

## Migration checklist

1. Create a Keirolabs key (1,000 free credits, no card).
2. Set `base_url="https://api.keirolabs.cloud/api/compat/exa/v1"` on `Exa(...)`.
3. Search your code for `find_similar`, `answer`, `type="deep*"`, `include_domains`, `exclude_domains`, `start_published_date`, `category`, `highlights` and `output_schema`. Keep those on Exa.
4. Pass `contents=False` and `type="keyword"` on searches that only need results; `exa-py` otherwise requests page text and caps you at 5 results.
5. Pass URLs, not Exa result IDs, to `get_contents`, one URL per call.
6. Run real traffic through both for a day and compare before switching fully.

## Verdict

Keirolabs is the cheapest way to move plain search and search-with-text off Exa without changing your client: $0.44 to $1.33 per 1,000 for results, $1.33 to $4.00 with page text, against $4 to $12 on Exa. Exa remains the better engine for semantic discovery, find-similar and its deep types. Most teams end up using both: Exa for discovery, a cheaper API for volume.

## FAQ

### What is the best Exa alternative?

For a drop-in switch, Keirolabs: it serves an Exa-compatible /search and /contents API, so you keep exa-py and set base_url. Search costs $0.44 to $1.33 per 1,000 on a plan against Exa's $4 to $7. Parallel suits low-latency agents, Tavily LangChain projects and Brave an independent index.

### How do I switch from Exa to Keirolabs?

Create a Keirolabs key, then construct Exa(api_key="keiro_...", base_url="https://api.keirolabs.cloud/api/compat/exa/v1"). Pass contents=False and type="keyword" for results-only searches, because exa-py requests page text by default.

### How much does the Exa API cost?

Exa charges $4 per 1,000 searches on lite and instant, $7 on auto and fast, $12 on deep-lite and deep, and $15 on deep-reasoning. Results past 10 add $1 per 1,000, and contents cost $1 per 1,000 pages for each type. New accounts get $10 a month free, no card.

### Which Exa features does the Keirolabs endpoint support?

Search (up to 35 results without contents, up to 5 with text or summary), includeEmbeddings, and /contents for one URL per call. Deep search types, find-similar, /answer, highlights, domain, date and category filters, and outputSchema are not supported yet.

### Is Keirolabs cheaper than Exa for page contents?

Not for a single content type. Exa charges $1 per 1,000 pages per content type; Keirolabs /contents costs $0.89 to $2.67 per 1,000 URLs. Keirolabs is cheaper on search and on search with page text in one call.

### Does Keirolabs support Exa's find-similar?

No. Find-similar and Exa's neural ranking have no equivalent yet. Keep those calls on Exa and move plain search and page fetches to a cheaper API.

### Is there a free Exa alternative?

Keirolabs gives 1,000 credits at signup with no card. Tavily gives 1,000 credits a month and Serper 2,500 queries once. Exa's own $10 monthly credit covers about 2,500 lite searches.

### Exa vs Parallel: which is better?

Parallel is cheaper and faster: $1 per 1,000 on turbo at about 200 ms. Exa is better for semantic discovery and find-similar. Both charge $1 per 1,000 for each result past 10.

## Related

- [Tavily alternative](https://keirolabs.cloud/tavily-alternative): Tavily-compatible endpoints at Keirolabs prices.
- [Parallel vs Exa vs Tavily vs Perplexity](https://keirolabs.cloud/parallel-vs-exa-vs-tavily): The four agent search APIs on price, speed and accuracy.
- [Exa API pricing](https://keirolabs.cloud/blog/exa-api-pricing-2026): Every Exa search type and content charge explained.
- [Exa alternatives (2026)](https://keirolabs.cloud/blog/exa-alternatives-2026): More replacements compared.
- [Exa vs Tavily](https://keirolabs.cloud/blog/exa-vs-tavily): The two most-used agent search APIs, side by side.
