Access Tiers

MidOS is a knowledge infrastructure for AI agents. Every query returns validated, battle-tested knowledge — not hallucinations, not outdated docs, not Stack Overflow guesswork.

Why MidOS Exists

Your AI agent is only as good as the knowledge it has access to. Without a curated, validated knowledge base:

• Tokens wasted: Your agent burns context window re-reading docs, retrying patterns, and debugging hallucinated code
• Time lost: Every “let me search for that” is 5-30 minutes of your day gone
• Quality unknown: When your agent says “use this pattern”, you have no idea if it’s correct, outdated, or dangerous

MidOS fixes this with a 5-layer validation pipeline. If it’s in EUREKA, it works. If it’s in SOTA, it’s the best known approach. Period.

What’s Inside

Asset	Quantity	What it means
Knowledge Chunks	21,000+	Distilled insights from 376+ books, talks, and papers
Executable Skills	118	Production-ready patterns across 20+ stacks
EUREKA Discoveries	168	Peer-reviewed, tested improvements
SOTA Entries	32	Best known solution for each problem
Truth Documents	29	Corrections to common mistakes and verified patches
AOTC Program	Active	Ahead-of-the-curve research (Ops/Enterprise exclusive)
MCP Tools	34+	Search, retrieve, analyze, coordinate — all via MCP
Vector Embeddings	290K+	Gemini 3072-d semantic search across the full library
Tech Stacks	20+	React, Next.js, FastAPI, Go, Rust, K8s, PostgreSQL, and more

Building this from scratch would take a team years. You get it instantly.

With MidOS vs Without

Scenario	Without MidOS	With MidOS
Agent needs a pattern	Hallucinates from training data. You debug for 30 min.	Gets EUREKA-validated code. Works first time.
New framework/stack	Half a day reading docs, tutorials, trial and error	get_skill("nextjs") — production-ready in seconds
Complex debugging	Hours of Stack Overflow + GitHub issues + guessing	hybrid_search("error description") — instant match
Token consumption	Agent re-reads entire docs every session. Burns 50K+ tokens on context.	Pre-distilled chunks. 500 tokens per answer instead of 5,000.
Knowledge freshness	Training cutoff was months ago. Docs may have changed.	Continuously harvested, validated, updated. 24/7.
Confidence level	”Is this correct?” — nobody knows	5-layer pipeline: if it passed EUREKA review, it’s correct
Team knowledge	Tribal knowledge lost in Slack threads	Shared episodic memory — what one agent learns, all agents access
Research depth	Manual googling. You do the work.	Automated research with validated sources, delivered as actionable briefs
Security patterns	Copy from blog posts. Hope it’s not vulnerable.	Audited security packs: OAuth, prompt injection, tool poisoning defenses
Cost per answer	Free (but your time isn’t)	$0.001/query but saves $10+/query in developer time

The economics are simple: MidOS costs less than one hour of developer time per month. It saves dozens of hours.

Tier Comparison

Feature	Community	Dev	Ops	Teams	Enterprise
Daily cap	50/day	500/day	2,000/day	500/day/IP	Unlimited
Rate limit	5/min	30/min	60/min	30/min/IP	100/min
IPs	1	1	1	12	100
Research	—	3/mo	30/mo (agent-driven)	50/mo (shared)	Unlimited (PenguChat)
Tools	9	34+	34+	34+	34+ custom
Knowledge	Truncated	Full 21K+	Full + AOTC	Full 21K+	Full + custom packs
Skills	Truncated	118 skills	118 + packs	118 skills	118 + custom
EUREKA + SOTA	—	168 + 32	168 + 32	168 + 32	Full
Truth	—	29	29	29	Full
AOTC	—	—	Exclusive	—	Full
Semantic Search	—	✓	✓	✓	✓
Security + Infra Packs	—	—	✓	—	✓
Episodic Memory	—	Search	Search + Write	Shared R/W	Shared R/W
Orchestration	—	Status	Pool signals	Pool signals	Full engine
Support	Community	Email	Priority (24h)	Email	Dedicated SLA
On-Premise	—	—	—	—	✓

Community — Evaluate, No Strings

No credit card. No expiration. No tricks. Permanent free access.

9 tools included:

• search_knowledge — keyword search across the full library
• list_skills — browse 118 skills, filter by stack
• hive_status / project_status — live system dashboard
• agent_handshake — personalized onboarding for your AI agent
• context_health — check your context window usage
• get_progress / where_was_i — session resume

Limitations (by design):

• Knowledge and skills truncated — enough to evaluate relevance, not for production
• No EUREKA, SOTA, Truth, or semantic search
• 50 queries/day, 5/min
• No research access

Who it’s for: Evaluating MidOS. Testing if it fits your workflow. Occasional use.

Dev — The EUREKA Line

Where MidOS becomes a force multiplier. Full knowledge, semantic search, and research briefs.

Everything in Community, plus:

Category	What you unlock
Full Knowledge	Untruncated 21,000+ chunks — complete code, explanations, context
Full Skills	118 executable patterns across 20+ stacks
EUREKA + SOTA	168 validated improvements + 32 best-known solutions
Truth	29 verified corrections and patches
Semantic Search	Hybrid search: keyword + vector + reranking
Research	3 briefs/month — submit a topic, get validated findings back
All 34+ Tools	Full tool access, no restrictions

ROI: If MidOS saves you 15 min/day, that’s 15x the subscription value at $50/hr.

Ops — Operation Specialists

For power users and agent operators. Exclusive content, agent-driven research, and orchestration.

Everything in Dev, plus:

Category	What you unlock
AOTC Inventions	3 original techniques — exclusive to Ops/Enterprise
Agent Research	30/month — agents trigger research autonomously
Security Packs	Tool poisoning, OAuth, prompt injection, MCP hardening
Infra Packs	Kubernetes, Terraform, observability, CI/CD configs
Episodic Write	Store agent experiences for reuse across sessions
Pool Coordination	Coordinate multiple agents via signals
Priority Support	Response within 24 hours

For comparison: GitHub Copilot Business ($19/user) gives autocomplete only. Cursor Pro ($20/mo) is IDE-locked. MidOS Ops gives a validated library + research + orchestration, compatible with ANY MCP client.

Teams — Scale to 12

Dev-tier access for your whole team. Shared memory and multi-agent orchestration.

Everything in Dev, scaled:

Category	What you unlock
12 IPs	12 machines/agents connected simultaneously
50 Research/month	Shared pool — any team member or agent can request
Shared Episodic Memory	What one agent learns, all agents access
Multi-Agent Orchestration	Pool coordination across the team
Team Dashboard	Usage visibility across all IPs

Who it’s for: Small teams (2-12 devs), agencies, startups using AI agents daily.

Note: Teams does not include AOTC or premium packs. For premium knowledge, upgrade to Ops individually or contact us for a custom bundle.

Enterprise — Organizational Scale + PenguChat

Full Ops access for 100 IPs. Unlimited research via PenguChat. On-premise available.

Everything in Ops, at scale:

Category	What you unlock
100 IPs	Organization-wide deployment
Unlimited Queries	No daily cap, 100/min — automated pipelines welcome
PenguChat Research	Conversational research intake — describe what you need, PenguChat validates and queues
Custom Knowledge Packs	Built specifically for your organization’s domain
On-Premise	Run MidOS inside your infrastructure
Dedicated SLA	Named support contact, guaranteed response times
Data Isolation	Private instance, no shared infrastructure

Research: The Core Differentiator

MidOS isn’t just a knowledge base — it’s a research engine. Here’s what “research” means at each tier:

Tier	Research Access	How it works
Community	None	—
Dev	3/month	Submit a topic via MCP or dashboard. Receive a structured brief within 24-48h.
Ops	30/month	Agent-driven: your agents can trigger research_request programmatically. Briefs include actionable code and validated sources.
Teams	50/month (shared)	Same as Dev, pooled across team. Any member can submit.
Enterprise	Unlimited via PenguChat	Conversational intake. No queue limit. Priority processing.

What a research brief contains:

• Problem statement and context
• Validated findings from MidOS knowledge + fresh sources
• Code samples (tested, not hallucinated)
• Trade-off analysis
• Recommended approach with rationale

Skills Coverage (20+ Tech Stacks)

Every skill is a complete, executable pattern — not a snippet. Includes context, trade-offs, and production-ready code.

Stack	Skills	Examples
Rust	3+	Ownership, Tokio, Axum, PyO3
Go	3+	Concurrency, error handling, gRPC
Security	4+	OAuth2, MCP hardening, prompt injection
Kubernetes	3+	Deployments, services, security
DevOps / CI/CD	5+	Terraform, Docker, Actions, Tekton, Argo
PostgreSQL	3+	Indexing, JSONB, full-text search
AI/ML	5+	RAG, LoRA, Gemini, Claude, LLM Guard
Data	4+	dbt, Iceberg, Feature Stores, Elasticsearch
Testing	3+	Vitest, Playwright, E2E
FastAPI / Python	4+	Async, DI, Pydantic v2, middleware
React / Next.js	5+	Server Components, hooks, App Router, migration
And more…	50+	Django, Laravel, NestJS, MongoDB, Redis, DDD

Moderator (Earned, $0)

Active community contributors earn Dev-tier access for free. Contribute quality knowledge, report issues, or help other users. If you make MidOS better, MidOS is free for you.

Try Now

# Connect via HTTP (any MCP client)
https://mcp.midos.dev/mcp

# Or install locally
pip install midos-mcp

No sign-up required for Community tier. Start querying immediately.

Payment

Auto-detected by region:

Region	Processor	Methods
Global (US/EU/Asia)	Paddle	Cards, PayPal, Apple Pay, Google Pay, Wire
LATAM (UY/AR/BR)	MercadoPago	Abitab, RedPagos, OCA, cuotas, QR
Crypto (optional)	Binance Pay	USDT, BTC, ETH

Paddle acts as Merchant of Record — taxes handled automatically. Annual plans save ~20%. 72-hour money-back guarantee on all tiers.

API Key Format

Tier	Key Prefix	Example
Dev	midos_sk_dev_	midos_sk_dev_abc123...
Ops	midos_sk_ops_	midos_sk_ops_xyz789...
Teams	midos_sk_team_	midos_sk_team_abc123...
Enterprise	midos_sk_ent_	midos_sk_ent_000...

export MIDOS_API_KEY="midos_sk_dev_your_key_here"