Direct programmatic AI query engine. Zero authentication required, silent key rotation, 5 core intelligence models, live DuckDuckGo + Wikipedia search, and drop-in OpenAI SDK compatibility.
Live Interactive API ConsoleDirect Gateway
Test the live KirkAI endpoints directly in your browser. Select a target model or task preset, provide a prompt, and observe the immediate response.
Ready. Click "Execute" to run an API query against the live Cloudflare deployment.
1. How the API System Works
Zero Friction & Free
No account creation, bearer tokens, or API paywalls are required. All requests automatically utilize the backend's shared community key pool with automated rotation and failover.
Dual Input Flexibility
Supports both ultra-simple queries (?prompt=... or {"prompt": "..."}) and complete OpenAI chat message arrays ({"messages": [{"role":"user", "content":"..."}]}).
Live Internet & Fact-Checking
Equipped with native real-time search (DuckDuckGo + Wikipedia) and live URL web scrapers. Fresh internet facts and verified citations are directly injected into model prompts.
2. Core Model Endpoints (/api/model/:model)
KirkAI features 5 dedicated models, ranked below from highest to lowest intelligence tier. Queries can be issued via GET /api/model/:model?prompt=... or POST /api/model/:model.
Lucent 4.0 Flagship (Default)
https://kirkai.pages.dev/api/model/lucent
Engine: space-bunny-alpha • 65k context
Most efficient kirk model, best for large scale database changes, project changes, or pretty much any useful need within an IDE. (Aliases: /api/model/stealth, /api/model/lucent-4.0).
Axiom 3.5 Refactor & Development
https://kirkai.pages.dev/api/model/axiom
Engine: nemotron-3-super-120b • 131k context
Great for major project changes, refactoring, major patch fixes, and overall development. (Aliases: /api/model/axiom-3.5).
Arco 3.0 Detail & Small Apps
https://kirkai.pages.dev/api/model/arco
Engine: qwen3.8-27b • 32k context
Quick eye for small details, efficient for small scale projects and applications.
Nova 2.3 Codebase & Quick Fixes
https://kirkai.pages.dev/api/model/nova
Engine: gemma-4-26b-a4b-it • 131k context
Efficient for small tasks, quick error fixes, general codebase analyzing.
Epoch 1.1 Fast Thinker & Search
https://kirkai.pages.dev/api/model/epoch
Engine: gemma-4-31b-it • 131k context
Fast thinker, methodical and analytical, perfect for quick web searches or questions.
3. Specialized Task Endpoints (/api/task/:task)
Instead of selecting a model manually, call a task endpoint. The API automatically sets the optimal model, applies tuned system instructions, and toggles real-time search.
Endpoint
Underlying Model
Special Capabilities
Ideal Use Case
/api/task/research
Nova 2.3
Live DuckDuckGo + Wikipedia Search
Real-time facts, current news, cited web sources
/api/task/code
Axiom 3.5 (120B)
Principal Software Engineer Persona
Production-ready code, architecture, unit testing
/api/task/reasoning
Epoch 1.1
Structured <thinking> Tags
Math, logic deductions, multi-step problem solving
/api/task/creative
Lucent (Stealth)
Expressive Literary Stylist
Storytelling, copy ideation, metaphoric writing
/api/task/stealth
Lucent (Stealth)
Direct Unfiltered Completion
Concise answers without unsolicited moralizing
/api/task/summarize
Nova 2.3
Actionable Key-Takeaway Extractor
Condensing long articles, documents, notes
/api/task/translate
Arco 3.0
Multilingual Polyglot
High-fidelity multi-language translation
4. Programmatic Friendlies Collaborative Engine
KirkAI includes a multi-agent pipeline where three autonomous agents critique and synthesize each other's outputs sequentially.
1. Erika The Organizer
Receives the user's premise, identifies core requirements, and outlines a structured initial draft.
2. Tyler The Detail Man
Audits Erika's draft with high-precision engineering scrutiny, flagging inconsistencies, gaps, and improvements.
3. Kirk The Decider
Synthesizes both perspectives and makes definitive, authoritative decisions to produce a complete deliverable.
# Execute the Friendlies collaborative pipeline
curl -X POST "https://kirkai.pages.dev/api/friendlies" \
-H "Content-Type: application/json" \
-d '{"prompt": "Architect a real-time multiplayer card game in Rust and WebSockets"}'
5. Drop-in OpenAI SDK Integration
KirkAI exposes /v1/chat/completions and /v1/models, allowing instant plug-and-play with existing AI client libraries.
from openai import OpenAI
# Connect to KirkAI with standard OpenAI SDK (no key needed!)
client = OpenAI(
base_url="https://kirkai.pages.dev/v1",
api_key="free-kirk-access" # Any string works
)
response = client.chat.completions.create(
model="lucent", # 'lucent', 'epoch', 'nova', 'arco', or 'axiom'
messages=[
{"role": "system", "content": "You are a software architect."},
{"role": "user", "content": "Design an event-driven pub/sub system."}
]
)
print(response.choices[0].message.content)
6. Code Integration Examples
# 1. Quick GET Query to Specific Model (ideal for quick scripts)
curl "https://kirkai.pages.dev/api/model/lucent?prompt=What+is+the+speed+of+light"
# 2. Raw Plaintext Output (ideal for terminal pipes)
curl "https://kirkai.pages.dev/api/model/lucent?prompt=What+is+the+speed+of+light&format=text"
# 3. POST Query with Web Research
curl -X POST "https://kirkai.pages.dev/api/task/research" \
-H "Content-Type: application/json" \
-d '{"prompt": "Latest discoveries from James Webb Space Telescope"}'
# 4. POST Query with Code Engineering
curl -X POST "https://kirkai.pages.dev/api/task/code" \
-H "Content-Type: application/json" \
-d '{"prompt": "Write an LRU Cache in Python with O(1) ops"}'
import requests
# Query specific model
res = requests.post(
"https://kirkai.pages.dev/api/model/lucent",
json={"prompt": "Explain distributed consensus in 3 sentences."}
)
data = res.json()
print("Model:", data["model"])
print("Reply:", data["response"])
# Query research task (includes verified sources)
res_search = requests.post(
"https://kirkai.pages.dev/api/task/research",
json={"prompt": "Who won Super Bowl 2024?"}
)
search_data = res_search.json()
print("Reply:", search_data["response"])
print("Sources:", search_data.get("sources", []))
// Vanilla JavaScript Fetch
async function queryKirkAI() {
const response = await fetch('https://kirkai.pages.dev/api/model/lucent', {
method: 'POST',
headers: { 'Content-Type': 'application/json' },
body: JSON.stringify({
prompt: 'Write a high-performance debounce function in TypeScript',
web_search: false
})
});
const data = await response.json();
console.log('AI Response:', data.response);
}
queryKirkAI();