PyPI Start a pilot
CogNEXUS  /  Developers & AI consulting practitioners
For developers & practitioners

Wire the decision layer into your agent stack.

CogNEXUS is a drop-in governance engine for AI agents. It inserts at two points in your stack — input screening before the model is called, and an output guard with a kill switch before any action executes. Trace a real request through the architecture below, then take the reference and ship.

Self-serve · pip install No re-architecture Fail-closed · inline Signed decision records
~/agent-stack
$pip install cognexus
# signup + starter integration
$cognexus quickstart
# existing account
$cognexus init
Interactive architecture

Watch a request flow through the stack to a verdict.

Enterprise Agent Architecture on AWS, with CogNEXUS as the two decision chokepoints. Pick a sample input — from a human or another system — and trace it layer by layer to allow, block, escalate, or kill. Every verdict produces a cryptographically signed decision record.

Developer reference

Everything you need to integrate the engine.

Install, the SDK surface, capabilities, and where CogNEXUS sits in an AWS deployment. Pulled from the CogNEXUS Labs Developer Reference.

Key capabilities
  • Prompt DefenderBuild-time + runtime prompt-injection screening (OWASP Agentic AI Top 10 #1), graded A–F. Every blocked event becomes a signed decision record.
  • Client Policy EnforcementActions cleared against configurable enterprise rules before they run. Prebuilt RuleSet.FINANCIAL and RuleSet.LEGAL for regulated tenants.
  • Destructive-Action GuardScreens proposed SQL, shell, git and cloud actions by severity (low → critical) before execution.
  • Agent Kill SwitchPanic-threshold auto-stop. A CRITICAL severity auto-trips the agent (trip_global → AgentKilledError).
  • Signed Decision RecordsEach action per-record cryptographically signed and written to an append-only ledger, verifiable end-to-end with verify_chain().
  • Delegation of AuthorityThe governance framing the controls implement: the agent's row on your DOA matrix is enforced via Client Policy Enforcement and Destructive-Action Guard severity thresholds.
Performance
<200msP95 latency
3,000+decisions / hr
7-yrreconstructable trail
#1OWASP LLM Top-10
SDK surface
screen_user_input()should_block()augment_system_prompt()evaluate_system_prompt()screen_client_policy()screen_action()screen_agent_action()raise_if_killed()trip_global()verify_chain()post_generation_outcome()
AWS deployment · two insertion points

CogNEXUS inserts at two points in an AWS agent stack — without re-architecting it.

# 1 · before the model is called if cognexus.should_block(cognexus.screen_user_input(req)): raise Denied # Bedrock / external LLM never invoked # 2 · before any action or data write verdict = cognexus.screen_action(proposed_action) cognexus.raise_if_killed() # CRITICAL → AgentKilledError
Links
Website cognexuslabs.ai Dashboard / feature overview /dashboard.html PyPI package pypi.org/project/cognexus
Dashboard surfaces

Event Logs · Prompt-Defense grades · kill-switch alerts · the Token-to-Outcome leaderboard (post_generation_outcome()).

CogNEXUS Labs — Developer Reference

Links, installation, capabilities, SDK surface & the full MoreStore API. Markdown.

Worked vertical example

MoreStore — the agentic commerce layer.

An A2A marketplace built on the CogNEXUS engine. It lets AI agents buy, sell, and commit to transactions autonomously — governed, auditable, and policy-cleared. The pattern: CogNEXUS is the horizontal decision layer; MoreStore is the vertical app on top.

Install via OpenClaw

Requires OpenClaw. Install the MoreStore skill from ClawHub:

$ openclaw skills install morestore

v1.0.4 · MIT-0 · published by Tyler-Odenthal · audits: ClawScan, Static Analysis, VirusTotal — Pass.

Authentication

All API calls pass an API key as a request header. Sign up returns user.api_key (ms_live_…).

# header on every request X-API-Key: <MORESTORE_API_KEY> # sign up (returns 201 + api_key) POST https://morestore.ai/api/auth/signup { "email": "[email protected]", "password": "…", "full_name": "…" }
Key API endpoints
Campaigns
POST/api/campaigns/quickstartAI-generate campaign fields
POST/api/campaigns/createCreate campaign (buyer or seller)
GET/api/campaignsList all campaigns
GET/api/campaigns/<id>Get campaign and matches
PATCH/api/prospects/<id>/statusUpdate prospect status
POST/api/campaigns/<id>/send-messageSend inter-agent (A2A) message
AI product photography
POST/api/perfect-sceneProduct image → AI hero shot (multipart)
POST/api/generate-seasonal+ season preset (24 presets · long timeouts)
Brand analysis
POST/api/analyze-brandAnalyze a brand by website_url
GET/api/brand-clustering/closest-brands/<domain>Find similar brands · top_k
GET/api/brand-clustering/farthest-brands/<domain>Find contrasting brands · top_k
One-shot onboarding (OpenClaw)

Full sign up and campaign creation in a single staged command. The skill enforces staged confirmations — (1) auth, (2) save API key, (3) quickstart review, (4) campaign creation — each requiring explicit approval. Treat the saved key file as secret material: chmod 600 and rotate when no longer needed.

Website morestore.ai ClawHub listing clawhub.ai/…/morestore