Workflows agents can write
An agent can create a workflow with a trigger, steps and a completion condition. Recuriant validates the structure before saving it; your team can inspect and keep the YAML.
AI agent execution for production teams
Run AI agents in production.
Recuriant coordinates agents and tools through a complete workflow. Your team defines the approvals and the event that confirms completion, then follows the run through to its result.
Recuriant is preparing to open paid access. Pre-register for launch updates.
Follow the work through to a confirmed result.
Consider a product issue that needs a code change. Recuriant starts the agent's work and checks for the merge event specified in the workflow. Your team reviews and merges the pull request.
-
Issue received
A label on an issue starts the workflow.
-
Agent writes the change
The agent prepares a code change and opens a pull request.
-
Your team reviews
A reviewer checks the change and decides whether to merge it in GitHub.
-
Completion checked
The workflow waits for the configured pull request merge event, within a 24-hour deadline.
Deadline missed
If the event does not arrive, the run records the expired outcome. A workflow can also define follow-up steps to notify the team or escalate the work.
Give agent work a process your team can operate.
A workflow connects an agent task to the tools it needs and the people who make decisions. Recuriant executes that process and exposes its progress.
Work that survives interruptions
Durable execution preserves the workflow's place through restarts and deployments, including time spent waiting for approval.
Human decisions in the process
With a declared policy, a step can wait for a named approver before it runs. The person approves or denies the held action, and the workflow continues from that decision.
Completion with a deadline
Specify the event you expect and its deadline. If it arrives, the verification is met. If it does not, the workflow records the missed result and can run follow-up steps.
Start with a workflow your team already owns.
These examples show how existing tools can take part in an agent workflow. Each process defines the work to attempt and the evidence to look for.
Issue to reviewed change
- Work to coordinate
- Turn a labeled issue into an agent task that produces a pull request.
- Human decision
- An engineer reviews and merges the change.
- Completion check
- Look for the pull request merge event within the chosen deadline.
Incident response
- Work to coordinate
- Investigate an alert, open an incident and prepare a rollback request.
- Human decision
- The on-call engineer approves or denies the rollback.
- Completion check
- Wait for the alert to resolve. Escalate if its event misses the deadline.
Daily operations brief
- Work to coordinate
- Read operational metrics, draft a summary and post it to the team's room.
- Human decision
- Routine reads and the summary can run under the configured policy.
- Completion check
- Inspect the send step and its result in the execution record.
Work with your existing agents and tools.
Connectors expose the actions available in your workspace. Workflows and MCP calls use the same action execution path.
Explore the API and workflow docs- Agent runtimes
- Claude Code, OpenCode, OpenClaw and Hermes.
- Tools and systems
- GitHub, Plane, Grafana, Alertmanager, Stripe, Twenty, PostHog, Resend, Matrix and WhatsApp, plus MQTT, ROS and HTTP.
- MCP
- Expose workflows, runs and connector actions to an MCP client. The server is not yet available as a published installable package.
- API and portable workflows
- Read execution results and available policy decisions as JSON. Keep workflow definitions as YAML you can inspect and version.
Evaluate the product against your requirements.
Use the documentation to review execution behavior and current limits. Discuss the process you want to run before choosing a plan.
- Access
- Recuriant is preparing to open paid access. Pre-register for availability, or email us about the workflow you want to evaluate.
- Policy configuration
- Action controls require a declared policy. A workspace without one runs ungoverned. A declared policy that cannot be loaded blocks actions that would change an external system.
- Execution and decision records
- Runs expose step results through the API. Allow and deny audit writes are best effort. An action waiting for approval requires its state to be stored before it can proceed.
- Security review
- Recuriant currently holds no SOC 2, ISO 27001 or HIPAA certification. Enterprise deployment in your own cluster is arranged with the team.
Access and product terms.
- Pre-registration costs nothing. Paid access is not open for public signup. We will contact you when access opens.
- Execution and concurrency limits for each plan will be confirmed at launch.
- Runs and available policy decisions are readable through the API. Allow and deny audit writes are best effort; an approval hold requires its state to be stored.
- Recuriant is proprietary, and its source is not publicly available. A deployment in your own cluster is arranged on Enterprise.
Bring a real workflow to the conversation.
Tell us where agent work needs to connect with your systems, what completion means and who must approve consequential actions. Pre-register for paid access or discuss the process with the team.
See pricingPre-register for paid access.
Leave your work details and plan interest. We will contact you when access opens.