Docs center for getting one real request working first.
Use this documentation center the way real users work: start with one successful API call, move into client setup, then choose your coding or agent workflow.
Start here
If a new user only needs the shortest path, send them through quick start, API reference and client config first.
Start with one successful test
These tools reduce the most common setup mistakes before a user commits to full client setup.
API tester
Verify Base URL, key, model name, and OpenAI-compatible responses before full configuration.
Open tool →Config generator
Generate cleaner config snippets for Hermes, OpenClaw, Codex, Cline, Roo Code, and more.
Open tool →Model selector
Pick a more suitable model based on coding, agent, speed, or budget priorities.
Open tool →Client and error guides
Client selection, model choice and common API errors for OpenAI-compatible workflows.
OpenAI-compatible clients that can connect to MonkeyAI.
A practical list of OpenAI-compatible clients and tools you can connect to MonkeyAI, including coding agents, desktop apps and S
Open guideHow to choose the best model for AI coding agents.
How to choose a model for AI coding agents by comparing reasoning quality, speed, context length, tool behavior and visible pric
Open guideFix API 401 invalid token errors.
Fix 401 invalid token errors in MonkeyAI and OpenAI-compatible clients by checking Authorization headers, API keys, domains and
Open guideFix HTTP 429 rate limits
Separate request limits from credit errors, honor Retry-After and keep SDK retries from multiplying traffic.
Open guideOpenAI Node.js SDK with a custom Base URL
Configure baseURL, model checks, endpoint parsing, retries and request IDs with the official JavaScript SDK.
Open guideOpenAI Python SDK with a custom Base URL
Set base_url, verify model IDs, handle timeouts and keep request IDs for practical debugging.
Open guideAnthropic Python SDK with a custom Base URL
Set base_url, x-api-key + anthropic-version, /v1/messages 401-auth-first probe, and a real client.messages.create() walkthrough on MonkeyAI.
Open guideAnthropic Node SDK with a custom baseURL
Use @anthropic-ai/sdk with baseURL on MonkeyAI; same wire protocol, same x-api-key + anthropic-version headers, Node streaming and error pattern.
Open guideCherry Studio and Chatbox with a MonkeyAI API key.
Configure Cherry Studio and Chatbox as OpenAI-compatible providers against MonkeyAI: API Host, API Key, model dropdown, vision input, streaming, and common 401 / 404 mistakes. Verified 2026-08-09.
Open guideResponses API vs Chat Completions
Compare endpoint fields, output parsing, conversation state and compatibility checks before migrating.
Open guideOpenAI-compatible streaming with SSE
Parse data frames, detect [DONE], distinguish HTTP errors from stream errors and debug stalled output.
Open guideSDK timeouts, retries and request IDs
Use bounded retries, set deliberate timeout values and preserve request IDs before escalating failures.
Open guideFix 503 “no available accounts” on an OpenAI-compatible API.
Distinguish the upstream account-pool 503 from 401/429/504: read the relay’s account_select_failed log line (captured on monkeytheme.com at 2026-08-08), then apply a 6-step caller recovery checklist.
Open guideFix ChatGPT Plus 5-hour OAuth pool exhaustion 503.
Diagnose the ChatGPT Plus OAuth rolling 5-hour window that hits zero on monkeytheme.com: read upstream_failover_switching and account_select_failed log lines (1,205 503s + 241 upstream 401s in 7 days, captured 2026-08-08), then drop concurrency and wait for the rolling window to recover.
Open guideFix API 504 timeout and upstream errors.
Understand 504 timeout and upstream errors in OpenAI-compatible API calls, and learn what to retry, what to log and when to cont
Open guideFix API 502 Bad Gateway and upstream bad-reply errors.
Distinguish HTTP 502 Bad Gateway from 504 and 503 by latency, upstream_failover_switching log shape, and avoid retry loops that re-burn the same broken upstream account.
Open guideChat Completions function calling on an OpenAI-compatible gateway
Wire tools[].function, read tool_calls, send role:tool messages, choose tool_choice values, and keep the executor sandboxed before any real call.
Open guideResponses API function calling on an OpenAI-compatible gateway
Define strict-JSON tools, parse call_id, return function_call_output, and reuse the executor pattern from Chat Completions without mixing endpoint shapes.
Open guideOpenAI-compatible embeddings API
POST /v1/embeddings with model ids pulled from /v1/models, Base URL https://monkeytheme.com/v1, and 401/404 probes before any application code.
Open guideOpenAI-compatible endpoint coverage on MonkeyAI.
Measured /v1 endpoint matrix on 2026-08-09: which OpenAI-style paths return 401 (auth-first, supported) and which return 404 page not found (not exposed). Includes a 3-row auth-header probe set.
Open guideDecision and troubleshooting guides
Pages for developer evaluation, pricing comparison, Hermes/OpenClaw setup and model-name errors.
AI API gateway for developers
A developer-focused guide to using MonkeyAI as an OpenAI-compatible gateway for SDKs and coding agents: pre-flight checklist, 6-step setup, real 401/404 probes on 2026-08-08, and operational habits for production apps.
Open guideAI API pricing comparison
Live GPT, Claude and Gemini prices on MonkeyAI: visible Codex/OpenAI rows, three Claude pricing tiers, how the percent-off badge is calculated, and what to compare before picking a model. Captured 2026-08-08.
Open guideHermes and OpenClaw provider setup with MonkeyAI: five-step safe order.
Set up Hermes Agent or OpenClaw with MonkeyAI by running five checks first: route, credential, config file, restart, then one short request — verified on monkeytheme.com on 2026-08-08.
Open guideFix model not found errors in OpenAI-compatible APIs
List models with the same key, copy an exact ID, then check Base URL and endpoint compatibility.
Open guideClient setup guides
Guides for high-intent setup searches: Codex CLI, Claude Code, Cline, Roo Code and Base URL /v1 issues.
Codex CLI API key setup with an OpenAI-compatible endpoint.
Set up Codex CLI with a MonkeyAI API key, OpenAI-compatible Base URL and one direct /v1/models test before changing your Codex con
Open guideClaude Code and API gateways
Understand how to use MonkeyAI around Claude Code, where OpenAI-compatible APIs fit, and what protocol boundaries to check before
Open guideCline and Roo Code setup with an OpenAI-compatible API.
Configure Cline or Roo Code with a MonkeyAI OpenAI-compatible API endpoint, Base URL, model name and direct validation steps.
Open guideWhy the API Base URL usually needs /v1.
Learn why OpenAI-compatible API clients use a /v1 Base URL, when to include it, and how to avoid duplicate or missing /v1 endpoint
Open guideSearch-intent guides
These pages answer the questions people usually search before they create an API key or configure a coding agent.
OpenAI-compatible API gateway
What compatibility means, when to use a gateway, and where /v1 fits.
Open guideGPT and Claude with one key
How one MonkeyAI key works across model families and coding-agent clients.
Open guideAPI key test and errors
Fix 401, 404, model not found and timeout issues before changing your client.
Open guideReading paths
Use a shorter path depending on what the visitor is actually trying to do.
By tool and workflow
When the user is coming from coding tools or agent clients, move them into the matching entry point instead of a generic landing page.
Agent client center
Hermes Agent, OpenClaw, Codex CLI, Claude Code, Cline and Roo Code guides on one page.
Open center →Coding tools
Setup notes for coding-oriented clients, protocol differences, and working examples.
Open guide →Agent tools
Use the API tester, config generator and model selector before asking users to guess.
Open tools →Coding tool setup guides
Setup guides for the coding agents and editor extensions people search for most: Hermes Agent, Codex CLI, Claude Code, and Cline or Roo Code.
Hermes Agent and OpenAI-compatible API
What order actually works when connecting Hermes Agent.
Open guide →Codex CLI and third-party API
How to switch Codex CLI without mixing version assumptions and config styles.
Open guide →Claude Code protocol boundary
What to say honestly when users ask about OpenAI-compatible support.
Open guide →Cline / Roo Code with OpenAI-compatible API
The cleanest setup path for extension-style coding clients.
Open guide →Billing and troubleshooting
Including the new Cryptomus top-up diagnosis page from 2026-08-08.
Fix Cryptomus USDT top-up not credited on monkeytheme.com.
Diagnose a Cryptomus top-up that stays PENDING, jumps to CANCELLED, or never credits the balance: official webhook path, 400 missing-sign signatures from real production logs, payment_orders reading order, and what evidence to send support.
Open guideKeep pricing and troubleshooting close to the setup path so users can self-serve instead of asking support first.
Billing and pricing
Understand token billing, balance, usage records and pricing behavior.
Open guide →FAQ troubleshooting
Fix 401, invalid model, quota, timeout and client-specific config mistakes.
Open guide →About, contact and how to buy
Three pages that used to live only in the footer/legal area are now reachable from the docs hub. They were already serving real, indexed English content; this round simply surfaces them.
About MonkeyAI
What MonkeyAI sells, who it is for, how it differs from selling bare OpenAI/Claude keys, and the team-vs-agent positioning story.
Open page →Contact MonkeyAI
Sales, billing, merchant review and API access questions, plus the public contact details and what evidence to send with each request type.
Open page →How to buy MonkeyAI
Step-by-step purchase path: register, choose plan or top up balance, generate an API key, and start calling /v1 — with the Cryptomus invoice resume_token recovery note.
Open page →Base URL
https://monkeytheme.com/v1