Agent orchestration
Use agents for judgment. Use workflows for control.
Combine durable, tool-using agents with explicit workflow graphs so every workload gets the right balance of autonomy, repeatability, and review.
One-time starter credit per account on Everstack Cloud. Expires 30 days after signup.
- 01Durable agent sessions
- 02Visual DAG workflows
- 03Parallel branches
- 04Human approval gates
01Operating model
One path from application intent to controlled execution.
Combine durable, tool-using agents with explicit workflow graphs so every workload gets the right balance of autonomy, repeatability, and review.
01
Choose the right execution model
Use a function for one bounded action, a workflow for explicit control flow, and an agent when the path depends on what the model discovers.
02
Compose without hiding the handoffs
Workflow nodes make model calls, agents, functions, HTTP requests, memory operations, and safety checks visible as one versioned graph.
03
Keep long-running work durable
Sessions, tools, approvals, streaming events, sandbox state, and subagent coordination remain managed by the runtime.
02Capabilities
The controls needed to run it in production.
Everstack keeps agents + workflows connected to the same identity, policy, deployment, and telemetry boundaries as the rest of the platform.
Tool-using agents
Run open-ended reasoning loops that can call tools, inspect results, and adapt their next action inside configured limits.
Visual workflow graphs
Connect typed nodes for models, agents, code, integrations, logic, safety, voice, and data.
Parallel execution
Run independent branches together when the dependency graph allows it.
Triggers
Start deployed workflows through webhooks, schedules, or direct API calls with explicit input variables.
Versioned deployment
Keep drafts separate from deployed workflow versions and roll back when an orchestration change misbehaves.
Approval boundaries
Pause sensitive tool calls or workflow steps until a reviewer explicitly approves the action.
03Where it fits
Built for real workloads, not isolated demos.
Start with the narrowest useful surface, then connect more of the platform as the workload needs stronger execution, policy, or evidence.
Adaptive operations
Let an agent investigate while a workflow controls the surrounding process.
Repeatable AI pipelines
Version explicit branches, retries, transforms, and review steps.
Background automation
Trigger durable work from schedules, webhooks, or application events.
04Questions
What teams usually ask first.
A direct view of how this surface behaves, where it runs, and how it connects to the rest of Everstack.
When should I use an agent instead of a workflow?
Use an agent when the model must choose the next action from intermediate results. Use a workflow when the steps and decision points should remain explicit.
Can a workflow call an agent?
Yes. Agent runs, functions, model calls, guardrails, integrations, and memory operations can participate in the same workflow graph.
Are workflow runs observable?
Yes. Execution events, node timing, payloads, token use, errors, and guardrail outcomes link into the platform trace.
Reference
The pages you open once you start building.
Continue through the stack