Put AI agents to work across your business without vendor lock-in. 🔒

Agent workflows made easy.

Organize your agents into "stacks" that share a common mission and watch them self-organize into dynamic and durable workflows automatically. Your agents will delegate tasks, hand off work, create their own scheduled tasks, form memories, and work around the clock to accomplish their mission.

• Bring people and personas together around a shared mission in Stack Stream.
• Watch your agents form a dynamic graph based on their responsibilities and mission.
• Monitor and interface with your agent team using the Stack Stream.

Monitor and optimize token spend across workspaces. 💰

Easily form AI agent teams that delegate less intensive reasoning tasks to more affordable models or to on-premises inference in your own data centers. Reserve higher-capability models for work that needs them.

• Monitor 24-hour and 30-day token use by persona and stack.
• Track token use across all workspaces in your organization, with stack and persona drilldowns.
• Set daily token limits for supported personas. Their runs pause when those limits are reached.

Easily change your AI provider at any time.

Configure shared AI providers for each workspace, then switch your agents between them while keeping their memory, integrations, mission, scheduled tasks, and other settings intact.

Experiment with models to find a more cost-efficient fit for each agent. Build teams with different models that automatically delegate work to the agents best suited to each task. Choose the providers that fit your needs without depending on a single provider.

• Change your agents' AI providers at any time without losing memory, scheduled tasks, integrations, skills, or run history.
• Each agent can have a different AI provider.
• Agents handoff work to teammates to balance speed, capability, and cost.
• Avoid AI provider lock-in. 🔒

AI Provider Agnostic

Run your  agents on your  hardware.

Deploy the PersonaStack Cluster Agent on your Kubernetes cluster to run agent pods on your own hardware. Connect your agent teams to internal networks to access local inference, source control, and other private systems when configured. Move agents between local hardware and PersonaStack hosting through provider settings. Moving execution locations can require new storage and re-enabling Dreaming.

• Install cluster-agent to link your Kubernetes clusters to PersonaStack.
• Agent pods can run on your Kubernetes clusters to reach internal services.
• Agent pods only run when there is work to do.
• Connect to local Ollama or OpenAI-compatible inference endpoints to run your own models.

Cluster Agent

Restrict agent access to only what is needed.

Take control of your agents' access and capabilities. Set integrations and capabilities before a run. Kubernetes worker runs execute in isolated pods. LAN access is configurable on supported AI provider targets.

• Restrict each agent to only the integrations and capabilities it requires.
• Enable access to local network resources on supported AI provider targets.
• Choose whether a supported worker persona keeps files between runs.

Collaborate with workspaces.

Provision shared workspaces for your employees to collaborate on building and operating AI agents and their teams. Share reusable stack templates across your organization. Collaborate on agent skills easily.

• Organize employee access using workspaces.
• Reuse AI provider configurations and integrations across all agents in a workspace.
• Build and manage shared stacks with your team.

Audit logging for humans and agents.

Review searchable agent runs and recorded administrative activity to understand what happened and identify follow-up.

• Keep audit logs of agent activity automatically.
• Allow agent auditors to monitor agent audit logs.
• Audit logs encrypted to your account.