1
Signup
Create an account and workspace. The first launch only needs email.
Unified LLM API with security and memory
The Unified Interface For Protected LLMs
One Spectorn key for model discovery, prepaid credits, fail-closed prompt checks, and conversation memory. Keep your OpenAI-compatible SDK.
300+
models in catalog
1 API
key for every call
Prepaid
spend control
Shield
scan before model
Claude Sonnet
reasoning
GPT-4o mini
fast general
Gemini Flash
long context
DeepSeek Chat
code analysis
Qwen Coder
multilingual
Llama 3.3
private route
One API for any model
Choose a model from the Spectorn catalog and call it through one OpenAI-compatible endpoint.
Browse allspectorn/auto
primaryfallbackprivate
Availability without rewrites
Routes and fallback live in the gateway while your client SDK and request shape stay the same.
Learn moreCredits12 450
Daily limit72%
Reservedsettled
Price and control
Prepaid balance, limits, keys, and usage history are collected in the user workspace.
View pricingprompt injectionchecked
data leakagechecked
policy routechecked
Protection by default
Prompt injection, jailbreak, leakage, and policy risk are checked before the request reaches a model.
View docsFeatured Models
A familiar catalog: model, use case, availability, and one Spectorn key.
Featured Workflows
The reasons teams buy more than a plain router: protected agents, MCP, memory, and security enablement.
$ spectorn guard run
repo context: scanned
tools: restricted
verdict: allow
Secure coding agents
MCP, tools, repo context and command guardrails for engineering teams.
RAGShieldMemoryAnswer
RAG and support bots
Memory, retrieval context and leakage checks for customer-facing assistants.
Bankingpolicy
Telecompolicy
Financepolicy
Banking and telecom
Threat libraries and policies for regulated AI deployments.
2
Buy credits
Add prepaid balance and keep team spend inside clear limits.
3
Get your API key
Point base_url at Spectorn and call models with the same SDK.
Spectorn adds the layer a plain catalog does not have
On the surface: models, keys, usage, and credits. Under each call: a protection gateway and optional persistent memory.
Security verdict
Requests are checked for prompt injection, jailbreak, exfiltration, and policy violations before the model call.
Conversation memory
Add persistent context with a session header, without a separate memory service.
Routing policy
Allow, deny, and reroute models by workspace, key, and task.
Audit trail
Keep verdict, usage, and route history for incident review and cost control.
drop-in · python
Drop-in integration
Only the endpoint changes. Your message format, SDK, and workflow stay familiar.
from openai import OpenAI
client = OpenAI(
base_url="https://spectorn.xyz/v1",
api_key="sk_spct_...",
)
resp = client.chat.completions.create(
model="spectorn/auto",
messages=[{"role": "user", "content": "Summarize this ticket."}],
extra_headers={"X-Spectorn-Session": "customer-42"},
)
# Spectorn checks the request before the model call.
# Output verdicts and usage are stored in your workspace.Security Library
A specialist knowledge base, from beginner onboarding to industry-specific LLM protection scenarios.
AI Threat Radar
Run LLM APIs through Spectorn
One key, familiar model catalog, credits, protection, and memory in one gateway.