Simple, Transparent Pricing
5 agents ship with the package. The Agent Store unlocks your full team. All three response tiers -- including the 35B model -- run locally at zero cost.
Open Source
5 agents built in. 7 more free from the Agent Store. 12 specialists, zero cost.
- [OK] 5 hardcover agents (orchestrator, marketing, content, analytics, python dev)
- [OK] 7 free Agent Store agents (web, JS, security, QA, docs, code archaeologist, setup guide)
- [OK] Compiled roundtables & contribution scoring
- [OK] Executive briefings (daily reports)
- [OK] Work queue orchestration
- [OK] Local inference (Ollama / llama.cpp)
- [OK] Persistent agent memory
- [OK] MCP server (tool integration)
- [OK] CLI (cohort command line)
- [OK] Community Discord
Pro
Full Agent Store access. 22+ specialist agents and premium pipelines.
- [OK] Everything in Open Source
- [+] 10+ additional agents (brand design, campaigns, email, social media, media production, hardware, databases, systems)
- [+] New agents as they ship to the Store
- [...] Website Creator pipeline (coming soon)
- [...] Content pipeline (RSS to publish) (coming soon)
- [OK] Cloud API override (Claude, GPT -- your key)
- [+] Priority email support
Enterprise
Custom agents built for your domain. Agent Factory, certification, and dedicated support. Currently in development -- contact us to discuss your needs.
- [OK] Everything in Pro
- [...] Agent Factory (create + train custom agents)
- [...] Agent Assessment & Certification
- [...] Supervisor Agent (autonomous oversight)
- [...] Advanced work queue (cron scheduling)
- [...] SSO / SAML authentication
- [...] Uptime SLA
- [...] Dedicated support engineer
- [...] On-premise / air-gapped deployment
The Smartest Tier Is Free
Most multi-agent platforms charge per token for their most capable model. Cohort's Smartest tier runs the 35B MoE model locally on dual-GPU systems -- 124 tok/s, no API key, no per-token cost. On a single RTX 3080 12GB, Smartest uses the full 9B model with extended thinking enabled. Cloud API is available as an optional override for users who want it, but it is never required.
What You're Actually Paying For
Cohort's core engine is MIT-licensed. The free tier ships 12 agents with full orchestration, executive briefings, and work queue. Pro unlocks your full team via the Agent Store and adds premium pipelines.
Cloud-based alternatives require per-token API costs that scale with agent count and usage. Cohort's local-first approach eliminates these ongoing costs.
Feature Comparison
| Open Source | Pro | Enterprise | |
|---|---|---|---|
| Hardcover agents | 5 | 5 | 5 |
| Agent Store agents | +7 free | +17 (full roster) | All + custom |
| Total agents | 12 | 22+ | Unlimited |
| Compiled roundtables | [OK] | [OK] | [OK] |
| Contribution scoring | [OK] | [OK] | [OK] |
| Work queue | [OK] | [OK] | [OK] |
| Executive briefings | [OK] | [OK] | [OK] |
| Persistent memory | [OK] | [OK] | [OK] |
| Local inference | [OK] | [OK] | [OK] |
| Website Creator | -- | Coming soon | Coming soon |
| Content pipeline | -- | Coming soon | Coming soon |
| Smart / Smarter / Smartest tiers | [OK] | [OK] | [OK] |
| Cloud API override (your key) | [OK] | [OK] | [OK] |
| Agent Factory | -- | -- | Planned |
| Agent Assessment | -- | -- | Planned |
| Cron scheduling | -- | -- | Planned |
| SSO / SAML | -- | -- | Planned |
| Uptime SLA | -- | -- | Planned |
| Support | Community | Priority email | Dedicated engineer |
Frequently Asked Questions
What is the Agent Store?
The Agent Store is Cohort's agent distribution service. When you run Cohort, it connects to the Store to fetch agent configurations, prompts, and learned expertise. Free users get 7 additional agents beyond the 5 built-in. Pro users get the full roster of 22+ agents, with new specialists added regularly. All agents run on YOUR hardware using local inference -- the Store just delivers the intelligence.
What hardware do I need?
A single consumer GPU with 8-12GB VRAM runs the default model (qwen3.5:9b) comfortably. An RTX 3060/3070/3080 or equivalent is plenty. CPU-only mode works but is slower. No cloud GPU required.
Is the free tier actually usable in production?
Yes. You get 12 agents -- 5 ship with the package, 7 more are free from the Agent Store. That covers orchestration, marketing, content strategy, analytics, Python, web, JavaScript, security, QA, docs, and code archaeology. Plus compiled roundtables, executive briefings, and work queue. The same engine runs across all tiers. Pro expands your team with specialist agents and adds premium pipelines like the Website Creator and Content Monitor.
Why are executive briefings free?
Executive briefings are how you see what your AI team accomplished. We want every Cohort user -- free or paid -- to start their day with a briefing. It's the best way to understand the value of multi-agent orchestration, and it naturally shows you what Pro and Enterprise capabilities could add to your workflow.
Can I use cloud models instead of local ones?
Yes, on all tiers. Cohort's Smartest tier runs locally by default (35B MoE on dual-GPU, 9B on single-GPU), but you can override any tier to use Claude, GPT, or any OpenAI-compatible API by adding your own API key in Settings. Built-in token budget limits prevent surprise bills.
What does "protocol-first" mean for vendor lock-in?
Cohort uses duck-typed Python protocols for agents, storage, and inference. Implement the interface and Cohort uses it. No base classes, no vendor SDK. pip install cohort has zero dependencies beyond the standard library. If you stop paying, your agents, data, and workflows remain fully functional on the open source tier.
When does Pro launch?
The Agent Store and Pro tier launch on March 22, 2026. Star the repo on GitHub to get notified. On launch day you'll be able to sign up for a free trial -- no credit card required.
Will you offer annual pricing?
Yes. Annual pricing details will be announced at launch. Enterprise pricing is negotiated per contract. Contact us for details.
Start Building Today
12 agents, zero cost. No signup, no API keys, no credit card. Just pip install cohort and go.