A new architecture for multi-agent systems _‍

The all-in-one agent platform that runs in your cloud.
Private. Secure. Built for teams who ship.

TRY PINAX
CONTACT US
Agent framework layer.
Top layer of a stacked isometric diagram with agents.
Top layer of a stacked isometric diagram with a memory card.
Top layer of a stacked isometric diagram with a grid matrix depicting the cloud.
Agent framework
Build self-learning agents with memory, knowledge, and guardrails. Any model. Any database. Your cloud.
Production runtime
Turn agents into a production service. Deploy anywhere. Ship on day one, not month six.
Built-in control plane
Chat, trace, and monitor from your browser. Your data stays in your system. No egress, no retention costs.
Secure and private by default
JWT, RBAC, and request-level isolation. Privacy and security are built into the architecture, not layered on.
Loved by engineers at
Tab 1
Tab 2
Tab 1
Tab 2
Tab 3
Tab 1
Tab 1
Tab 2
Tab 1
Tab 2
Tab 1
Tab 2
A new operating system is emerging.
Agents are no longer experiments. They’re infrastructure.
‍
They need more than a framework. They need a runtime, a control plane, and security that keeps data private.
‍
Pinax provides all three. _

Pinax runtime

Turn agents into production infrastructure. Run agents, teams, and workflows as one scalable API. Ship on day one.

more about Pinax
Pinax
Agent
Team
Workflow

agent_os = AgentOS(
 description ="Powerful Agent System",
 agents =[knowledge_agent, support_agent],
  teams =[research_team],
   workflows=[social_media_workflow],
   interfaces =[Slack(), AISdk(), AGUI()],
)

agent_os = AgentOS(
 description ="Powerful Agent System",
 agents =[knowledge_agent],
  teams =[research_team],
  workflows =[sm_workflow],
  interfaces =[Slack(), AISdk()],
)

1
2
3
4
5
6
7

knowledge_agent = Agent(
 name ="Knowledge Agent",
 model ="claude:sonnet-4",
  tools=[DeepResearchTool],
   knowledge =Knowledge("company_docs")
   db=Postgres(" postgresql://user:pass@host/db"),
   enable_memories=true
 instructions ="Search internal docs to answer questions",
)

knowledge_agent = Agent(
 name ="Knowledge Agent",
 model ="claude:sonnet-4",
  tools=[DeepResearchTool],
  knowledge=Knowledge("company_docs")
  db= Postgres(connection_string),
  enable_memories=true
 instructions =instruction,
)

1
2
3
4
5
6
7
8
9

research_team = Team(
 name ="Research Squad",
 members =[web_researcher, social_insights_agent],
 model ="claude:sonnet-4",
   db=Postgres("postgresql://user:pass@host/db"),
 instructions ="Collaborate for deep research",
 enable_memories =true,
)

research_team = Team(
 name ="Research Squad",
 members =[agent 1, agent 2],
 model ="claude:sonnet-4",
  db=Postgres(connection_string),
 instructions =instruction,
 enable_memories =true,
)

1
2
3
4
5
6
7
8

social_media_workflow = Workflow(
   name=Social Media Autopilot",
   description =description, 
   db=Postgres(connection_string),
   steps=[
       Router(
           selector =select_channel,
           choices=[agent 1, agent 2],
       ),
       publish_post,
   ],
)

1
2
3
4
5
6
7
8
9
10
11
12

social_media_workflow = Workflow(
   name=Social Media Autopilot",
   description ="Generate & publish engaging posts.", 
   db=Postgres(" postgresql://user:pass@host/db"),
   steps=[
       Router(
           selector=select_channel,
           choices =[x_agent, linkedin_agent],
       ),
       publish_post,
   ],
)

Pinax SDK

Build agents with memory, knowledge, tools, guardrails, and human-in-the-loop. One framework, everything included.

More about Framework
Instructions
Memory
Knowledge
Self Learning
Guardrails

Production-ready

Private by design

Security built-in

Scalable

Manage your system with a powerful control plane

A secure UI for your AgentOS. Full visibility and real-time control for engineers and operators. Chat, trace, monitor, and manage.

Track, evaluate and improve system performance
Edit, organize and label user memories
Add, update and manage knowledge used by your agents
In-depth insight into every live interaction
Evaluate your agents across 3 dimensions: accuracy, reliability and performance.
[ Metrics ]
[ Memory manager ]
[ Knowledge manager ]
[ Session monitoring ]
[ Evaluations ]

Performance matters_

Fastest agent instantiation

529×

faster than Langgraph

57×

faster than PydanticAI

70×

faster than CrewAI

Lowest memory footprint

24×

lower than Langgraph

4×

lower than PydanticAI

10×

lower than CrewAI

Bar chart comparing agent instantiation time: 3 μs (Pinax) vs 1178 μs (Status quo).Bar chart comparing agent instantiation time: 3 μs (Pinax) vs 1178 μs (Status quo).

Time to instantiate an agent (avg.)

Bar chart comparing memory footprint per agent: 6,656 bytes (Pinax) vs 136,649 bytes (Status quo).Bar chart comparing memory footprint per agent: 6,656 bytes (Pinax) vs 136,649 bytes (Status quo).

Memory footprint per agent (avg.)

Private by default. No data leaves your cloud.

Your AgentOS runs in your cloud. Usage, logs, metrics, traces, memory, knowledge, sessions, and user data stay in your environment remain fully under your control.

Diagram showing Pinax architecture where your app, Agent OS API, and control plane UI run fully in your own cloud, keeping all data private and under your control.An isometric diagram of 3 layers: your app, agent OS, your cloud
Diagram showing a non-Pinax setup where your app connects to your cloud, but AI backend, monitoring service, and monitoring UI are hosted externally, sending data outside your environment.An isometric diagram of 2 sets of layers. One with your cloud and your app and a second one linked to it with external cloud and a monitoring system
[ With Pinax ]
[ Without Pinax ]

Monitor system in real-time

Keep everything in your database

Any cloud: AWS, GCP, Railway

Get in touch_

Have questions or need help? Reach out to the Pinax team and we’ll get back to you.
support@pinax.tech

The future runs on Pinax_

Everything you need to build, run and manage secure multi-agent systems in your cloud.

Try Pinax