CertainLogic Stack
One install. Four components. Everything your AI needs to stop hallucinating, stop burning money, and start proving its work.
Founding Tranche — $29/mo
First 100 users or first 30 days. Lock in $29/mo forever. After founding tranche: $59/mo retail.
Your first month of token savings covers the year. 60-85% cost reduction is guaranteed in production.
What You Get
| Component | What It Does | Your ROI |
|---|---|---|
| Company Brain Core OS | Deterministic knowledge base. Instant verified answers. No hallucinations. | 122× faster responses, 98% hit rate on core facts |
| Token Reduction Engine (TRE) | Intelligent cache + LLM Intent Protection. Blocks dangerous queries and caches every safe answer. | 60-85% token cost reduction, 79.7% cache hit rate |
| AgentPathfinder | HMAC-SHA256 audit trails on every tool call. Catches false successes, hanging operations, and failures. | Compliance-ready. Proves what your agents actually did. |
| Smart Router (Coding + Intents) | Routes every query to the cheapest capable model. Coding → DeepSeek V4 Flash ($0.20/1K queries). Architecture → Kimi K2.7 Code. | 2ms routing decision. Never waste premium models on simple queries. |
One-Click Install
clawhub install certainlogic-stack && cd certainlogic-stack && bash setup.sh
The setup script installs all 4 components, generates your config, and runs verification. Takes ~30 seconds.
What setup.sh does
- Checks prerequisites (Python 3.8+, clawhub CLI)
- Installs all 4 components via
clawhub install - Generates
stack-config.json - Verifies everything installed correctly
- Prints quick-start commands
Manual Install (if you prefer)
clawhub install company-brain-os
clawhub install certainlogic-tre
clawhub install certainlogic-pathfinder
clawhub install smart-router-coding
clawhub install smart-router-intents
Quick Start
from company_brain import Brain
brain = Brain()
# Instant verified answers — no LLM call, no hallucination
result = brain.query("What is our refund policy?")
# Returns only verified facts
# With TRE cache (60-85% token savings on repeat queries)
from tre import get_cached_answer
result = get_cached_answer("Same question") # <10ms, $0
# Route a coding query to the cheapest capable model
openclaw run smart-router-coding "Write a binary search in Rust"
# DeepSeek V4 Flash: $0.0008/query
Architecture
User Query
↓
LLM Intent Protection (blocks dangerous/wasteful queries)
↓
Cache Check (TRE) → Hit = 8ms, $0, 0% hallucination
↓
Smart Router → cheapest capable model
↓
Company Brain (verified facts) or LLM
↓
Hallucination Guard (never stores uncertain output)
↓
AgentPathfinder Audit (HMAC-SHA256 signed)
↓
Response + Persistent Log
Production Metrics
| Metric | Value |
|---|---|
| Cache hit rate | 79.7% |
| Warm cache latency | 8ms |
| Token cost reduction | 60-85% |
| Intent classification | 94% accuracy at 2ms |
| Verified facts | 1,170+ |
| Audit trail integrity | HMAC-SHA256, tamper-evident |
Verification
After install, run:
cd certainlogic-stack && python3 test.py
Or check individual components:
python3 -c "from company_brain import Brain; b = Brain(); print(b.query('test'))"
python3 -c "from tre import get_cached_answer; print('TRE loaded')"
python3 -c "from agentpathfinder import ToolAuditor; print('Pathfinder loaded')"
Licensing
BSL 1.1 with startup grant — free for organizations ≤3 employees or ≤$100K annual revenue. Converts to Apache 2.0 on 2029-07-11.
See LICENSE file for full terms.
Support
Built with brutal honesty by CertainLogic — https://certainlogic.ai
Questions: anton@certainlogic.ai