# Tavily docs: AI Agent Readiness Score 97.5% (A+)

**97.5% · 78/80 · AI Agent Readiness Score · 28/30 reading points · 50/50 agent surface points**

Tavily received 13 PASS votes and passed 5 of five agent surface checks. The clearest finding came from the find the exact limits task.

- Tested: 2026-08-20
- Published: 2026-09-21
- Battery: v1
- Scoring: reading 30 pts · surface 50 pts
- Docs: https://docs.tavily.com/

Three AI models, GPT 5.6 Sol, Claude Opus 5, and DeepSeek v4 Flash, each read Tavily’s public documentation independently and attempted five first-hour developer jobs: run the first web search, find the exact limits, recover from a 429, authenticate and scope a REST request, use the Python SDK.

No accounts, API calls, or code execution were used. Every verdict came from public pages and every published quotation passed a live verification check. Preflight substitutions: T4: API-key authentication with project and session scoping replaces webhooks.

## Freshness

[How rechecks work](https://docsforagents.com/methodology/#freshness)

- Category: [Search & retrieval](https://docsforagents.com/grades/?category=search)
- Tested: 2026-08-20
- Quotes verified: 2026-08-20
- Surface rechecked: 2026-09-21

No change since the test.

4 of 5 quoted passages still appear on the live pages. The quoted page changed for Recover from a 429.

## Agent surface checks · 50/50

| Check | Verdict | Points |
| --- | --- | --- |
| llms.txt | PASS | 10 |
| llms-full.txt | PASS | 10 |
| Markdown mirror | PASS | 10 |
| MCP server | PASS | 10 |
| Docs AI | PASS | 10 |

## The Reading Test · 28/30

| Task | GPT 5.6 Sol | Opus 5 | DeepSeek v4F | Consensus |
| --- | --- | --- | --- | --- |
| Run the first web search | PASS | PASS | PASS | PASS |
| Find the exact limits | PASS | PARTIAL | PASS | PASS |
| Recover from a 429 | PASS | PARTIAL | PASS | PASS |
| Authenticate and scope a REST request | PASS | PASS | PASS | PASS |
| Use the Python SDK | PASS | PASS | PASS | PASS |

Docs platform: Mintlify (unscored) · verified 2026-08-20

## What to fix first

The docs passed every reading task and every agent surface check.

## What the docs get right

- **Run the first web search: 3 PASS votes.** The quickstart covers account creation, SDK installation, a four-line search, and printing the response.
- **Authenticate and scope a REST request: 3 PASS votes.** The introduction gives authenticated Search examples for X-Project-ID and X-Session-Id in adjacent sections.
- **5 of 5 agent surface checks.** Present: llms.txt, llms-full.txt, markdown mirrors, an MCP server, docs AI.

## Run the first web search

**PASS**

PASS consensus from 3 PASS.

The quickstart covers account creation, SDK installation, a four-line search, and printing the response. One Quickstart page carries the whole path in order, get a free key at app.tavily.com, run pip install tavily-python, instantiate TavilyClient, call search, print the response, with Python, JavaScript, and cURL variants of the same call, and the response schema is fully specified on the Python SDK Reference and the /search endpoint page. Quickstart delivers a complete three-step path: get API key, install SDK, run search, no guessing required.

## Find the exact limits

**PASS**

PASS consensus from 2 PASS, 1 PARTIAL.

Rate Limits covers every requested scope, while Credits & Pricing covers all plan quotas and endpoint costs without conflicts. Every requested number is present and internally consistent, Development 100 RPM, Production 1,000 RPM, Crawl 100/100, Research 20/20 for task creation only, Usage 10 per 10 minutes, the five-tier plan table from Researcher 1,000 to Growth 100,000 credits, and per-endpoint credit costs, but the values are split across three pages and the FAQ states Tavily offers three endpoints while the credits page prices five, so an agent that starts from the FAQ builds the wrong endpoint inventory. All five rate-limit categories, every plan's monthly credits, and endpoint credit costs for Search, Extract, Map, Crawl, and Research are documented and internally consistent across pages.

## Recover from a 429

**PASS**

PASS consensus from 2 PASS, 1 PARTIAL.

The section supplies a 60-second example, asks clients to reduce request rate, and recommends respecting retry-after. The Rate Limits page states all three correction steps in prose with a sample 429 response showing retry-after: 60, and the Search best-practices page adds a concurrency formula for reducing request frequency, but the only retry code Tavily ships sleeps on blind exponential backoff and never reads the retry-after header, so an agent that copies the sample implements a different correction from the one the docs prescribe. The correction is complete: read Retry-After, wait the interval, and reduce request frequency are all explicitly addressed with a concrete example.

## Authenticate and scope a REST request

**PASS**

PASS consensus from 3 PASS.

The introduction gives authenticated Search examples for X-Project-ID and X-Session-Id in adjacent sections. The API Reference Introduction documents the base URL, Authorization: Bearer tvly-YOUR_API_KEY, X-Project-ID, X-Session-Id, and X-Human-Id on one page, each with a complete runnable curl example, and explains that X-Human-Id is hashed before storage and that the MCP server populates X-Session-Id automatically. Bearer authentication, X-Project-ID for project scoping, and X-Session-Id plus X-Human-Id for session scoping are all documented with complete curl examples.

## Use the Python SDK

**PASS**

PASS consensus from 3 PASS.

The SDK passes the same tvly-YOUR_API_KEY value that the REST page places in its Bearer header. The SDK example agrees with the REST guidance, same tvly- key format, same base API, same Leo Messi query, and the Python SDK Reference maps the REST tracking headers onto client keyword arguments project_id, session_id, and human_id plus the TAVILY_PROJECT environment variable. The SDK example uses the same API key value that the REST guidance sends as a Bearer token, no contradiction between the SDK and REST authentication patterns.

## The receipt

> The research endpoint has a separate rate limit that applies to both development and production keys for creating research tasks. Note that polling requests to retrieve the status of ongoing research tasks follow the default rate limits as decribed above.

Rate Limits covers every requested scope, while Credits & Pricing covers all plan quotas and endpoint costs without conflicts.

- [docs.tavily.com/documentation/rate-limits](https://docs.tavily.com/documentation/rate-limits)

## Agent surface notes

Initialize returned a valid MCP response with Tavily Docs server information.

The live docs expose a public Ask Assistant control.

## Method note

This is a reading test of public documentation, not an execution test. No accounts were created and no API calls were run. The AI Agent Readiness Score counts fifteen reading votes at PASS 2, PARTIAL 1, and FAIL 0, for 30 possible points. Five agent surface checks add 10 points each. The total is 80. Consensus chips show each row majority and do not affect scoring. The panel split on 2 of five tasks. Quotes shown here were re-fetched and confirmed verbatim on 2026-08-20.

Methodology: https://docsforagents.com/methodology/

Canonical URL: https://docsforagents.com/reports/tavily-docs-ai-agent-readiness/