Published 2026-09-21 · Tested 2026-08-19
Deno
D+Deno received 9 PASS votes and passed 3 of five agent surface checks. The clearest finding came from the find the exact limits task.
Three AI models, GPT 5.6 Sol, Claude Opus 5, and DeepSeek v4 Flash, each read Deno’s public documentation independently and attempted five first-hour developer jobs: deploy the first application, find the exact limits, recover from a warmup timeout, verify a build event, use the TypeScript API.
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: A deployment warmup timeout replaced HTTP 429 because it is a documented migration error. T4: GitHub repository_dispatch build events replaced configurable outgoing webhooks.
Freshness
How rechecks work- Category
- Cloud & developer platforms
- Tested
- Quotes verified
- Surface rechecked
No change since the test.
4 of 5 quoted passages still appear on the live pages. The quoted page changed for Use the TypeScript API.
67.5% · 54/80 · AI Agent Readiness Score · reading 30 pts · surface 50 pts
Agent surface checks · 30/50
The Reading Test
| Task | GPT 5.6 Sol | Opus 5 | DeepSeek v4F | Consensus |
|---|---|---|---|---|
| Deploy the first application | PASS | PARTIAL | PASS | PASS |
| Find the exact limits | PARTIAL | PARTIAL | PARTIAL | PARTIAL |
| Recover from a warmup timeout | PASS | PASS | PASS | PASS |
| Verify a build event | PARTIAL | PARTIAL | PASS | PARTIAL |
| Use the TypeScript API | PASS | PASS | PASS | PASS |
docs platform: Custom Deno Deploy site (unscored) · verified 2026-08-19
What to fix first
These 4 fixes could add up to 25 points to the AI Agent Readiness Score. The list ranks each fix by the points it would add. How the ranking works
- 1 +10 points MCP server check failed
Found: Initialize returned HTTP 405, and no public endpoint completed initialization.
Fix: Run a public MCP server for the docs that answers a JSON-RPC initialize call and offers a docs search tool.
Evidence: docs.deno.com/
mcp (opens in a new tab) - 2 +10 points Docs AI check failed
Found: The live docs expose standard search but no public assistant control.
Fix: Add an assistant to the docs site that answers questions from the docs and links to its sources.
- 3
Found: The CLI reference allows 30-minute build timeouts, but the builds reference caps them at 15. In the panel's reading, the limits page gives a different runtime memory limit.
Fix: Use the same runtime memory and build timeout values on the limits page, the builds reference, and the CLI reference.
Evidence: docs.deno.com/
deploy/ reference/ builds (opens in a new tab) - 4
Found: The apps reference gives the build events and payload, but no expected sender or verification rule.
Fix: On the apps reference, state who can send the build repository_dispatch events and what a workflow can trust in client_payload.
Evidence: docs.deno.com/
deploy/ reference/ apps (opens in a new tab)
What the docs get right
- Recover from a warmup timeout: 3 PASS votes. The migration guide links the timeout to legacy serve() and covers entrypoint and dependency upgrades.
- Use the TypeScript API: 3 PASS votes. The minimum Deno.serve example returns a Response and matches the migration guide's required API.
- 3 of 5 agent surface checks. Present: llms.txt, llms-full.txt, markdown mirrors. Missing: an MCP server, docs AI.
Deploy the first application
PASSPASS consensus from 2 PASS, 1 PARTIAL.
The CLI tutorial creates code, authenticates, deploys it, and says the deployment returns a live URL. The getting-started guide walks the whole path from organization to first build, but it says "Mono-repos (repositories where the application lives in a subdirectory) are not yet supported" while the Builds reference documents an "App directory" option that is "Useful for monorepos", so an agent cannot tell whether a subdirectory app can be deployed. The path is unambiguous: create an organization, click '+ New App', configure the repo and build settings, and deploy. The getting-started guide and apps reference describe the same sequence of steps.
Find the exact limits
PARTIALPARTIAL consensus from 3 PARTIAL.
The limits page says 512MB, while the builds reference says 768MB by default and 4GB on Pro. The page filed under "Policies & Limits" contains no numeric quota, no overage rule, and a future-tense promise about a date ten months past, so plan quotas and overage pricing are only obtainable from the marketing pricing page outside the docs; build timeout, build memory, runtime memory, and regions are documented but the build-timeout maximum contradicts the CLI reference. Plan quotas live on the pricing page (deno.com/deploy/pricing not docs.deno.com). Runtime memory default 768 MB, max 4 GB on Pro. Build memory defaults to 3 GB, can increase to 4 GB on Pro. The builds reference page says build timeout max is 15 minutes on Pro, but the CLI reference page lists allowed values up to 30 minutes, a direct contradiction. Region counts are spread across three pages: the main Deploy page says '2 regions', the migration guide names 'us and eu', and the CLI allows 'us, eu, global'. No single page assembles all limits.
Recover from a warmup timeout
PASSPASS consensus from 3 PASS.
The migration guide links the timeout to legacy serve() and covers entrypoint and dependency upgrades. The migration guide names the symptom, the cause, the corrected code as a diff, and the third-party-dependency case, so diagnosis and repair need no cross-referencing, though the warmup timeout duration itself is never stated. The migration guide explicitly names the legacy std serve() as the cause, explains the warmup timeout symptom, and provides the exact fix: replace serve() from std/http with Deno.serve(). It even warns about transitive dependencies importing the old function. Recovery is complete from the docs alone.
Verify a build event
PARTIALPARTIAL consensus from 1 PASS, 2 PARTIAL.
The docs provide the trigger and payload, but no expected sender or verification rule for Deno-originated events. The four event names, the full TypeScript payload type, and a runnable GitHub Actions workflow are all on one page, but no Deno page states who is authorized to dispatch these events or whether client_payload can be forged by any other actor holding repository dispatch rights, so the trust boundary has to be established from GitHub's documentation instead. The apps reference documents all four event types, provides the full TypeScript payload interface with app, organization, and revision fields including git SHA and preview_url, and gives a complete GitHub Actions workflow example that receives deno_deploy.build.routed. The trust boundary is clear: events are sent by Deno Deploy to the linked repository.
Use the TypeScript API
PASSPASS consensus from 3 PASS.
The minimum Deno.serve example returns a Response and matches the migration guide's required API. The same minimal handler appears in the API reference, the HTTP server guide, and the migration diff, and the Deploy runtime reference confirms Deploy "starts the application using the configured entrypoint and waits for the HTTP server to start", so the example is deployable as written. The main docs page and the migration guide both use Deno.serve with a handler returning a Response object. The API reference confirms that Deno.ServeHandler consumes a Request and returns a Response. The example matches current deployment guidance on both the new Deploy platform and the runtime docs.
The receipt
Runtime memory limit (optional): The maximum amount of memory the application can use at runtime. Defaults to 768 MB, can be increased to 4 GB on the Pro plan.
The limits page says 512MB, while the builds reference says 768MB by default and 4GB on Pro.
Agent surface notes
Initialize returned HTTP 405, and no public endpoint completed initialization.
The live docs expose standard search but no public assistant control.
Show the score
Paste this into a readme:
[](https://docsforagents.com/reports/deno-docs-ai-agent-readiness/) 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 2026-08-19.
Put another docs site through the battery.
Nominate a docs site