Published 2026-08-20 · Tested 2026-08-19

E2B

B+

E2B received 9 PASS votes and passed 5 of five readiness checks. The clearest finding came from the recover from a 429 task.

Panel: GPT 5.6 Sol, Opus 5, DeepSeek v4F Battery: v1 Verified: 2026-08-19 Read as markdown (opens in a new tab)

Three AI models, GPT 5.6 Sol, Claude Opus 5, and DeepSeek v4 Flash, each read E2B’s public documentation independently and attempted five first-hour developer jobs: create a sandbox, find the exact limits, recover from a 429, verify a lifecycle webhook, use the TypeScript 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.

E2B Mintlify · published 2026-08-20
B+

87.5% · 70/80 · AI Agent Readiness Score · reading 30 pts · surface 50 pts

llms.txt PASS
llms-full.txt PASS
markdown mirror PASS
MCP server PASS
docs AI PASS
Task GPT 5.6 SolOpus 5DeepSeek v4F Consensus
Create a sandbox PASSPARTIALPASS PASS
Find the exact limits FAILPASSPASS PASS
Recover from a 429 FAILFAILFAIL FAIL
Verify a lifecycle webhook PASSPASSPASS PASS
Use the typescript sdk PASSPASSPARTIAL PASS

docs platform: Mintlify (unscored) · verified 2026-08-19

What the docs get right

  • Verify a lifecycle webhook: 3 PASS votes. One page documents registration, payload receipt, signature verification, retry triggers, and idempotent deduplication by event ID.
  • Create a sandbox: 2 PASS votes. The quickstart covers signup, API-key setup, installation, sandbox creation, hello-world execution, root listing, and the launch command.
  • 5 of 5 readiness checks. Present: llms.txt, llms-full.txt, markdown mirrors, an MCP server, docs AI.

Create a sandbox

PASS

PASS consensus from 2 PASS, 1 PARTIAL.

The quickstart covers signup, API-key setup, installation, sandbox creation, hello-world execution, root listing, and the launch command. The quickstart covers create, execute, and list in five steps, but it installs @e2b/code-interpreter and calls runCode while the docs home page, /api-key, and /commands all install e2b and call sandbox.commands.run, and no page explains which package to pick. The quickstart provides a clear 5-step path from account creation, API key setup, SDK install, code to run, and execution command with no missing steps.

Find the exact limits

PASS

PASS consensus from 2 PASS, 1 FAIL.

Pro resource ceilings use open-ended plus signs, while most Enterprise fields say Custom; exact cross-tier limits are unavailable. The Plans table on /billing states every requested dimension per tier (vCPUs 8 / 8+ / Custom, memory 8 GB / 8+ GB / Custom, disk 10 GB / 20+ GB / Custom, runtime 1 hour / 24 hours / Custom, concurrent sandboxes 20 / 100-1,100 / 1,100+, concurrent builds 20 / 20 / Custom, creation rate 1 per sec / 5 per sec / Custom) and the numbers are repeated consistently on /sandbox, /filesystem, /template/quickstart, and the FAQ pages. The billing page and pricing page together state all limits for each tier; values are broadly consistent across pages.

Recover from a 429

FAIL

FAIL consensus from 3 FAIL.

The reference names the cause but provides no retry timing, response-header guidance, backoff, or limit-remediation steps. Across all 241 documentation pages the only 429 material is this one-line class description (plus its Python twin and a passing mention that fork errors map 429 to RateLimitError); no page states which limit produces a 429, no Retry-After header is documented, and the words backoff and retry never appear in a rate-limit context, so there is no recovery path to follow. RateLimitError is documented as a class with a one-line description, but the docs give no recovery path: no retry strategy, no backoff guidance, no rate limit headers, and no connection between sandbox creation rate (1/sec Hobby, 5/sec Pro on the billing page) and the 429 error.

Verify a lifecycle webhook

PASS

PASS consensus from 3 PASS.

One page documents registration, payload receipt, signature verification, retry triggers, and idempotent deduplication by event ID. One page carries the whole job: registration with signatureSecret, the full JSON payload schema, working e2b-signature verification code in JavaScript, Python, and Go, all four e2b-* request headers, the retry count (up to 3 attempts total), what counts as a failed delivery, and the instruction to deduplicate on the payload id field. Registration, payload format, signature verification code (SHA-256(secret+payload) base64) in JS/Python/Go, retry policy (3 attempts, 4xx/5xx failure), and idempotency guidance are all documented on one page.

Use the typescript sdk

PASS

PASS consensus from 2 PASS, 1 PARTIAL.

The quickstart and SDK reference both import @e2b/code-interpreter and call Sandbox.create without arguments. The e2b package, the import { Sandbox } from 'e2b' minimum example, and the commands and files modules on the versioned v2.38.2 reference match the home page snippet and the /commands page exactly. The quickstart has a complete working example using @e2b/code-interpreter, but the lifecycle events API page imports { Sandbox } from 'e2b' , two different npm packages exposing the same API shape, creating confusion when cross-referencing pages.

The receipt

Thrown when the API rate limit is exceeded.

The reference names the cause but provides no retry timing, response-header guidance, backoff, or limit-remediation steps.

Agent surface notes

Initialize returned JSON-RPC protocol 2025-06-18 and server name E2B Docs.

The live docs expose assistant-entry controls labeled Toggle assistant panel.

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 readiness checks add 10 points each. The total is 80. Consensus chips show each row majority and do not affect scoring. The panel split on 3 of five tasks. Quotes shown here were re-fetched and confirmed verbatim on 2026-08-19.

Read the full methodology

Put another docs site through the battery.

Nominate a docs site