1K4.AI
Open Lab

AI. The Right Way.

One workspace. Every model. Your rules. Code, documents, presentations, deployment. No subscriptions ever.

Multi-model agents that peer-review each other.
Free Claude Code integration. Local models welcome.

Open Lab

YOU are in control.

See what 1K4 Lab can do.

CONVERSATION
I need to write an NSF research grant application for a climate modeling project
Agent is gathering requirements
What's your target funding amount?
$150K–$300K$300K–$500K$500K+
Do you have preliminary data?
Yes, publishedYes, unpublishedNot yet
OUTPUT
Structured Spec
Project Narrative
Multi-scale climate modeling with Arctic amplification focus. 3-year scope covering observational validation and predictive ensemble runs.
Budget Justification
2 postdoc positions, cloud compute allocation (150K GPU-hours), field station access fees, publication costs.
Timeline
Year 1: Data collection and model calibration. Year 2: Ensemble simulations. Year 3: Validation, publication, and dissemination.

Your code. Fully annotated.

Review findings become inline annotations. Select code, leave comments, submit batches. The agent reads your feedback and iterates.

1
2async def validate_token(token: str) -> bool:
3
4 try:
5 payload = jwt.decode(token, SECRET)
6 return payload["exp"] > time.time()
7
8 except jwt.DecodeError:
9 return False
10
11
12+ except jwt.ExpiredSignatureError:
13+ logger.warning("Expired token: %s", token[:8])
14+ return False
ANNOTATIONS
CRITICAL — Lines 4-6
Missing ExpiredSignatureError handler. Expired tokens fall through to generic DecodeError.
Agent applied fix (diff above)
Your comment — Line 5
Should we also validate the algorithm? jwt.decode without algorithms param is insecure.
2 comments queued
GITbranch: fix/token-validation+14 -2

Any model. One key.

Intelligent routing picks the right model for each task. Or choose your own. Or bring your local models at zero cost.

MODEL ROUTING RULES
Code→Claude Sonnet 4.5
Reasoning→Claude Sonnet 4.5
Image Parsing→GPT-4.1
Orchestrator→Gemini 2.5 Flash
Subagent Workers→Inherit from parent
Simple→Gemini 2.5 Flash
Reset to defaults
OR CHOOSE YOUR OWN
Anthropic
Claude Opus 4Claude Sonnet 4.5Claude Haiku 4.5
OpenAI
GPT-5.2GPT-5.2 Proo3o3 Pro
Google
Gemini 2.5 ProGemini 2.5 Flash
xAI
Grok 4Grok 3
$0
Bridge connected
Local (Ollama)
gemma4:26bllama3.3:70bmistral-large
npx @1key4ai/bridge@latest --api-key sk-xxx

Your hardware. Your data. Free.

MULTIMODALImageGPT-image-1, DALL-E 3VideoVeo 2AudioGPT-4o AudioVisionGPT-4.1, Gemini 2.5 Pro

Connect any app. Change one line.

OpenAI-compatible endpoint. Swap your API base URL, keep your workflow.

from openai import OpenAI
client = OpenAI(
api_key="sk-your-1k4-key",
base_url="https://api.1key4ai.com/v1" # ← that's it
)
# Use any model from any provider
response = client.chat.completions.create(
model="claude-sonnet-4-5", # or any model from /pricing
messages=[{"role": "user", "content": "Hello!"}]
)
# Or let routing pick the best model
response = client.chat.completions.create(
model="1k4-custom", # smart routing
messages=[{"role": "user", "content": prompt}]
)
Chat completionsFunction callingImage generationStreamingBearer auth

Tested with Cursor, TypingMind, Continue.dev, OpenWebUI, and any OpenAI SDK client.

∞

Smarter than a prompt.

1K4 agents don't just respond. They plan, coordinate, and adapt to the complexity of your work.

∞Infinite context

Your conversation never hits a wall. The agent manages its own context window and memory so nothing important gets lost, however long you chat. Work on complex projects without starting over. Bye-bye context window anxiety.

Parallel agents

Complex tasks get broken down and distributed across specialized agents working simultaneously. Each focused on what they do best, coordinated automatically.

Adapts to your task

The agent actively adjusts to the complexity of your task and to the level of help you need. Yes, even to a bad prompt.

Remembers your project

Context carries across sessions. The agent recalls past decisions, understands your codebase structure, and builds on previous work. No re-explaining every time.

Pay as you go. Zero subscriptions.

$1 = 1 credit. Never expires. Use what you need.
Stop when you want.

FREE

Your own models

Connect your local Ollama instance via 1K4 Bridge. Full agent capabilities on your own hardware.

$0

Performance depends on your hardware and model. Dedicated cloud inference coming soon.

Cloud models. Pricing that's hard to beat.

Buy credits. Use any model from any provider. Per token, per image, per second. No overage. No expiry.

Gemini 2.5 Flashfrom $0.00/M input
GPT-4.1 minifrom $0.00/M input
Claude 3.5 Haikufrom $0.00/M input
Full pricing table →Start with $5
View All Model Prices

Full usage control.

Create projects, pick providers, track spend. Everything in one place.

New Project

Endpoint

api.1key4ai.com/v1

API Key

1k4-proj_xxxx···xxxx

Custom Model

1k4-custom

Provider

Select provider...
40 models · 89ms
Your Usage at a Glance

$0.00

spent

0

requests

$0.000

avg/req

Projects will appear here

Questions.

How does pricing work?

No subscriptions ever. No commitment. $1 = 1 credit. Credits never expire. When you use a cloud model, credits are deducted based on that model’s token pricing (text) or per-unit pricing (images, video, audio). Local models via the 1K4 Bridge cost nothing. See our pricing page for exact rates.

No subscriptions ever. No commitment. $1 = 1 credit. Credits never expire. When you use a cloud model, credits are deducted based on that model’s token pricing (text) or per-unit pricing (images, video, audio). Local models via the 1K4 Bridge cost nothing. See our pricing page for exact rates.

Yes. You get free Lab messages, free API requests, and a free Claude Code plan review to try everything out. No credit card required. Pay only when you want more.

Lab is the 1K4 agent workspace. You chat with an AI agent that can actually do things: edit documents, write and review code, manage files, deploy to servers, coordinate parallel tasks. It’s not a chatbot with a text box. It’s a workspace with real execution capabilities.

Every major provider: OpenAI, Anthropic, Google, and xAI. Text, image, video, and audio models. We keep adding new models. Check our model list and product updates regularly to stay tuned.

Yes. Install Ollama on your machine, run npx @1key4ai/bridge@latest --api-key sk-xxx, and your local models appear in Studio with full agent capabilities. Zero credits consumed. Performance depends on your hardware.

A virtual model endpoint that routes each request to the best model based on your rules. Define presets: code goes to Claude, reasoning goes to o3, simple queries go to Flash. All connected apps benefit instantly. Routing happens in under 5ms.

1K4 provides a multi-model plan reviewer for Claude Code via MCP. Run npx -y -p @1key4ai/mcp-review@latest mcp-review setup --api-key YOUR_API_KEY, and Gemini, GPT, and Grok review your plans before you implement. Catches architecture gaps, missed edge cases, and wrong assumptions. Set up in under a minute. More Claude Code integrations coming soon for power users.

Yes. The agent provisions real Linux servers, installs dependencies, deploys your code, configures domains, and sets up routing. You get a live URL, not a preview that vanishes. Sandbox instances are perfect for prototyping and staging: spin up, test, tear down. Persistent compute tiers for production hosting are coming soon. All managed by the agent, no DevOps required.

Absolutely. Change your API base URL to api.1key4ai.com and use your 1K4 API key. That’s it. Works with any OpenAI-compatible application. Same SDK, same code, different endpoint. You can create as many API profiles as you wish to connect multiple apps. You can change the model routing for any API profile at any time.

API Proxy mode is stateless: we forward your request and forget it. No content stored. Studio mode stores conversations and files to power the workspace; this data is encrypted at rest and in transit, accessible only through your authenticated session. In both modes, your data is never used for model training.

API requests return an error. No overage charges, ever. You will never be billed beyond what you’ve prepaid. Add more credits to resume. Local models via the Bridge are unaffected since they don’t use credits.

Ready to try?

Get Started

Free to start. No credit card required.