Turn scattered company knowledge into AI-ready skills.
YCB (Your Company Brain) connects to 38 enterprise tools, automatically chunking unstructured messages into machine-readable procedure cards that AI agents execute, and answers any question directly in your terminal.
“Refunds under $100 within 30 days don’t need finance approval.”
Single combined prompt classifies type & extracts key concept arrays.
Handle Customer Refunds & Disputes
- • Verify Stripe ID
- • Check 30-day window
Experience the YCB workflow in real time.
YCB lives in your terminal. Ask natural-language questions, switch local or cloud AI models, and inspect auto-generated procedure cards.
📋 Handle Customer Refunds & Disputes
- Verify transaction ID in Stripe / Billing portal against HubSpot deal record.
- Check customer account age — if under 30 days, approve standard refund without manager sign-off.
- If subscription is active, prompt user whether to cancel renewal or issue partial credit.
- Trigger automated confirmation receipt via Slack (#billing-ops) and customer email.
Documentation eats hours. AI agents still don't know how your company operates.
Institutional knowledge is scattered across thousands of Slack messages, Notion pages, and GitHub PRs. Company Brain transforms unstructured chaos into executable intelligence.
AI agents give generic, hallucinated answers
Internal chatbots have no memory of team decisions, internal protocols, or system architecture. Employees re-explain context every single session.
Manual runbooks go stale the day after they're written
Engineers spend hours authoring Notion wikis that nobody maintains when services, payment gateways, or deployment scripts change.
Knowledge is locked in silos across 38 tools
Critical edge cases live in a 6-month-old Slack thread, while the refund policy is split between HubSpot deal notes and Stripe docs.
Agents receive strict, executable Skill cards
AutoGen, LangChain, and internal bots plug into skills_file.json containing explicit step-by-step procedures and if/then rules.
Continuous background synthesis & extraction
Syncs automatically every 30 minutes, chunking new conversations and updating procedure cards without manual human intervention.
Unified 38-connector offline knowledge base
A single natural-language query (ycb --ask) searches across Slack, GitHub, Linear, Jira, and Google Workspace in 300ms.
Machine-readable procedure cards for AI agents.
Instead of raw messy text dumps, Company Brain extracts actionable step-by-step procedures with strict domain schema enforcement.
Handle Customer Refunds & Billing Disputes
Standardized procedure for evaluating customer refund eligibility, processing Stripe charge adjustments, and recording customer retention notes.
- ✓Stripe Billing Admin or HubSpot Support Specialist permissions
- ✓Access to #billing-ops Slack channel for manager sign-offs
Step 1: Lookup and verify transaction
Locate customer Stripe Customer ID (cus_xxx) via email in HubSpot. Cross-reference invoice charge ID and verify card authorization status.
Step 2: Check 30-day freshness window
If transaction was billed within the last 30 calendar days, proceed directly without escalation. If > 30 days, tag @finance-lead in #billing-ops.
Step 3: Issue credit or cash reversal
Execute refund in Stripe Dashboard with reason "Requested by customer". If customer wishes to stay on free tier, downgrade plan immediately.
Step 4: Send confirmation & update ticket
Dispatch automated receipt, log resolution in HubSpot ticket, and mark conversation closed with tag "refund_processed".
How YCB works
Connect across 38 tools
YCB hooks into Slack, Notion, GitHub, Google Workspace, Linear, and Jira via unified BaseConnector handlers with rate-limit backoff.
Idempotent batch ingestion
Pulls messages, pull requests, docs, and incident threads in structured batches with cryptographic deduplication. It never processes the same item twice.
Local AI model inference
Runs Gemma 4 locally via Ollama on localhost:11434. Your raw messages and confidential discussions never leave your internal infrastructure.
Single-call concept extraction
Simultaneously classifies content type (procedure, policy, decision, incident) and extracts key concepts, halving API latency with heuristic confidence scoring.
Synthesize Pydantic Skill cards
Related concepts are clustered into domain models with explicit step-by-step procedures, prerequisites, if/then decision rules, and edge cases.
Export & terminal Q&A
Outputs clean skills_file.json for autonomous AI agents (LangChain, AutoGen) and enables multi-turn terminal natural-language search with ycb --ask.
The 5-stage knowledge synthesis pipeline.
Engineered from the ground up for privacy, idempotency, and high-accuracy extraction across 38 enterprise platforms.
Ingestion & 38 Connectors
Pulls raw documents, messages, issues, and customer tickets through a unified BaseConnector interface with error isolation.
- →Abstract BaseConnector enforces ingest(days_back=30) and retry_with_backoff() for 429/5xx rate limits.
- →Dynamic registry.py loads configured connectors automatically without hardcoding.
- →Idempotency check filters out already processed items via get_processed_source_item_ids().
- →One failing connector logs an error and continues — never blocking the rest of the sync run.
class SlackConnector(BaseConnector):
def ingest(self, days_back: int = 30) -> List[RawDataItem]:
# Idempotent fetch with exponential backoff
for channel in self.get_joined_channels():
messages = self.retry_with_backoff(
lambda: self.client.conversations_history(channel=channel.id)
)
for msg in messages:
yield RawDataItem(
id=f"slack:{channel.id}:{msg['ts']}",
source=DataSourceType.SLACK,
content=msg['text']
)Connects to your entire company stack.
From Slack and Google Docs to GitHub, Linear, Jira, HubSpot, and Rippling. Ingest communication, engineering, support, and HR knowledge in one place.
Your company data never leaves your hardware.
Confidential Slack threads, salary discussions, security post-mortems, and customer PII should never be transmitted to third-party cloud AI vendors.
Company Brain executes inference locally via Ollama (Gemma 4), stores raw data and chunks in an on-premise SQLite database, and enforces strict offline idempotency.
data/company_brain.dboutput/skills_file.jsonBuilt for speed, accuracy, and absolute sovereignty.
Rigorous benchmarks measured across multi-connector syncs, sentence chunking, and local LLM synthesis in YCB.
Documentation Time
Procedures that took 45 minutes to draft manually are synthesized and formatted in under 2 minutes.
Enterprise Sources
Unified across communication, code, tracking, HR, customer support, and analytics tools.
Deduplicated Sync
Running sync 10 times produces the exact same result as running once. Resume interrupted syncs instantly.
Local Sovereignty
Inference (Gemma 4 via Ollama), ingestion, chunking, and SQLite storage all execute on your own servers.
Designed for teams that value speed and data ownership.
Whether you're an engineer building autonomous workflows or a COO streamlining operations.
Fast-Growing Startups & Scaleups
High onboarding velocity, rapidly evolving product specs, and thin documentation. Transform informal Slack consensus into official SOPs automatically.
AI Agent & Tool Builders
Engineers building autonomous agents with LangChain, AutoGen, or Cursor who need structured company procedures with exact decision trees.
Regulated & Security-Conscious Orgs
FinTech, healthcare, and enterprise teams with strict data residency requirements that cannot send internal documents to public cloud AI APIs.
Distributed Teams Across 38+ SaaS Apps
Teams tired of hunting across Slack threads, Notion wikis, Linear backlogs, and Jira boards just to find out how a simple process works.
Start offline for free. Unlock all 38 connectors when ready.
Instant self-serve license key issuance. Deploy on your local terminal with 100% offline sovereignty.
Solo Tier
Essential local knowledge synthesis for individual engineers and founders.
- ✓2 Basic Connectors (Google Docs + Slack)
- ✓100% offline local inference (Gemma 4 via Ollama)
- ✓Terminal natural-language search (
ycb --ask) - ✓Single-machine hardware binding (1 seat)
- ✕36 remaining enterprise connectors
Max Tier
Full enterprise connector suite with multi-provider model routing and priority chunking.
- ✓All 38 Connectors Unlocked (GitHub, Notion, Jira, Linear, etc.)
- ✓Multi-provider model switcher (Claude, OpenAI, OpenRouter)
- ✓Idempotent batch sync & deduplication
- ✓Unlimited generated Skill cards (
skills_file.json) - ✓Instant Razorpay activation
Enterprise Fleet
For security-conscious organizations deploying across multi-tenant air-gapped clusters.
- ✓All 38 connectors with OAuth 2.0 & SAML SSO
- ✓Dedicated private GPU cluster deployment
- ✓Custom connector engineering & data retention policies
- ✓Multi-seat licensing with manual provisioning
- ✓99.9% uptime SLA & dedicated support engineer
Need to evaluate full 38-connector capabilities?
Run it locally or schedule a 15-min walkthrough.
Install the open-source CLI directly, or drop your work email for a tailored walkthrough reading your team's Slack and Notion offline.