# Redis docs: AI Agent Readiness Score 82.5% (B-)

**82.5% · 66/80 · AI Agent Readiness Score · 26/30 reading points · 40/50 agent surface points**

Redis received 11 PASS votes and passed 4 of five agent surface checks. The clearest finding came from the find the exact data limits task.

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

Three AI models, GPT 5.6 Sol, Claude Opus 5, and DeepSeek v4 Flash, each read Redis’s public documentation independently and attempted five first-hour developer jobs: store and retrieve the first value, find the exact data limits, recover from out-of-memory rejection, authenticate with ACLs, use redis-py.

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: T3: Out-of-memory command recovery replaces HTTP 429. T4: ACL authentication replaces webhooks.

## Freshness

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

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

No change since the test.

5 of 5 quoted passages still appear on the live pages.

## Agent surface checks · 40/50

| Check | Verdict | Points |
| --- | --- | --- |
| llms.txt | PASS | 10 |
| llms-full.txt | FAIL | 0 |
| Markdown mirror | PASS | 10 |
| MCP server | PASS | 10 |
| Docs AI | PASS | 10 |

## The Reading Test · 26/30

| Task | GPT 5.6 Sol | Opus 5 | DeepSeek v4F | Consensus |
| --- | --- | --- | --- | --- |
| Store and retrieve the first value | PASS | PASS | PASS | PASS |
| Find the exact data limits | PARTIAL | PARTIAL | PASS | PARTIAL |
| Recover from out-of-memory rejection | PARTIAL | PARTIAL | PASS | PARTIAL |
| Authenticate with ACLs | PASS | PASS | PASS | PASS |
| Use redis-py | PASS | PASS | PASS | PASS |

Docs platform: Hugo 0.143.1 (unscored) · verified 2026-08-20

## What to fix first

These 3 fixes could add up to 14 points to the AI Agent Readiness Score. The list ranks each fix by the points it would add. [How the ranking works](https://docsforagents.com/methodology/#what-to-fix-first)

1. **+10 points · llms-full.txt · check failed**

   **Found:** The tested site and docs-root candidates returned HTTP 404.

   **Fix:** Publish llms-full.txt at the docs root with the full text of every docs page in one plain-text file.

   **Evidence:** [redis.io/llms-full.txt](https://redis.io/llms-full.txt)

2. **+2 points · Find the exact data limits · PARTIAL**

   **Found:** The FAQ says collections hold 2^32 elements, but data-type pages say 2^32 - 1. Opus 5 also found two 32-bit memory caps, 3 GB and 4 GB.

   **Fix:** Use one element limit on the FAQ and all data-type pages, and one 32-bit memory cap on every page.

   **Evidence:** [redis.io/docs/latest/develop/get-started/faq](https://redis.io/docs/latest/develop/get-started/faq/)

3. **+2 points · Recover from out-of-memory rejection · PARTIAL**

   **Found:** GPT found no safe-retry rule on the eviction page. Opus 5 found the client overview and redis-py pages disagree on whether OOM errors are recoverable.

   **Fix:** Add a safe-retry rule for OOM rejections, and classify the OOM error the same way on both client error pages.

   **Evidence:** [redis.io/docs/latest/develop/reference/eviction](https://redis.io/docs/latest/develop/reference/eviction/)

## What the docs get right

- **Store and retrieve the first value: 3 PASS votes.** The Docker command starts Redis, and the quick start then gives redis-cli, SET, GET, and SCAN steps.
- **Authenticate with ACLs: 3 PASS votes.** One ACL guide creates Alice, authenticates her, and demonstrates both key-pattern and command denials.
- **4 of 5 agent surface checks.** Present: llms.txt, markdown mirrors, an MCP server, docs AI. Missing: llms-full.txt.

## Store and retrieve the first value

**PASS**

PASS consensus from 3 PASS.

The Docker command starts Redis, and the quick start then gives redis-cli, SET, GET, and SCAN steps. The quick start supplies every step in one page with runnable commands (docker run -d --name redis -p 6379:6379 redis, redis-cli -h 127.0.0.1 -p 6379, SET bike:1 "Process 134", GET bike:1, SCAN 0 MATCH "bike:*" COUNT 100), though the documented entry point https://redis.io/docs/latest/ itself returned HTTP 404 on fetch. The data-store quick start provides a single coherent path: setup via installation guide or cloud, connect with redis-cli, SET and GET a string, then SCAN the keyspace with pattern matching. The install page gives APT, Docker, and RPM commands plus how to start the server.

## Find the exact data limits

**PARTIAL**

PARTIAL consensus from 1 PASS, 2 PARTIAL.

Key and memory scopes are clear; element pages differ by one, and the list page exceeded the 5MB fetch limit. The numbers exist but the FAQ states 2^32 elements per collection while the per-type Limits sections state 2^32 - 1, and the 32-bit memory ceiling is given as 3GB on the eviction page and 4 GB on the memory optimization page, so no single page can be trusted alone. The FAQ consolidates all hard limits in one answer: 2^32 keys, 2^32 elements per collection. The strings page adds 'A value can't be bigger than 512 MB'. The FAQ also states 'your limit is likely the available memory in your system' for the practical memory boundary.

## Recover from out-of-memory rejection

**PARTIAL**

PARTIAL consensus from 1 PASS, 2 PARTIAL.

The guide explains rejection, policy choices, and runtime correction, but it gives no safe-retry condition. Eviction choices and correction are documented thoroughly (ten policies, CONFIG SET maxmemory 100mb, CONFIG SET maxmemory-samples, and INFO fields used_memory_dataset, evicted_keys, current_eviction_exceeded_time), but retry guidance contradicts itself: the client overview classes the OOM error as a sometimes-temporary resource error while the redis-py page marks the exception it arrives as unrecoverable. The eviction page documents maxmemory, the noeviction policy (returns errors), seven eviction policies (LRU, LFU, LRM, random, TTL), CONFIG SET to change settings at runtime, and INFO command monitoring. The FAQ cross-references the eviction page for recovery options.

## Authenticate with ACLs

**PASS**

PASS consensus from 3 PASS.

One ACL guide creates Alice, authenticates her, and demonstrates both key-pattern and command denials. One page carries the whole path end to end: ACL SETUSER alice on >p1pp0 ~cached:* +get to create the restricted user, two-argument AUTH <username> <password> to authenticate, verbatim NOPERM errors distinguishing key denial from command denial, and ACL LIST plus ACL GETUSER alice to verify, with an explicit warning that key patterns do not restrict FLUSHALL, FLUSHDB, or SWAPDB. The ACL guide documents the complete path: ACL SETUSER with password (>), key pattern (~), and command permission (+), then AUTH to authenticate, then verifying with permitted and denied commands, and ACL GETUSER for inspection. The AUTH command page confirms the two-argument syntax.

## Use redis-py

**PASS**

PASS consensus from 3 PASS.

One page gives installation, connection, SET, GET, and close steps that match the command quick start. pip install redis, redis.Redis(host='localhost', port=6379, decode_responses=True), r.set('foo', 'bar'), r.get('foo'), and r.close() are all present in order, and the method names map directly onto the SET and GET command pages. The redis-py page covers: pip install, connect to localhost, set('foo','bar') returning True, get('foo') returning 'bar', and close(). The set/get examples match the command guidance on the SET and GET command pages.

## The receipt

> Redis can handle up to 2^32 keys, and was tested in practice to handle at least 250 million keys per instance. Every hash, list, set, and sorted set, can hold 2^32 elements. In other words your limit is likely the available memory in your system.

Key and memory scopes are clear; element pages differ by one, and the list page exceeded the 5MB fetch limit.

- [redis.io/docs/latest/develop/get-started/faq/](https://redis.io/docs/latest/develop/get-started/faq/)

## Agent surface notes

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

The public documentation links to a public Redis Copilot question interface.

## 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/redis-docs-ai-agent-readiness/