Connect anything with an API
Import a definition or describe an endpoint. Every endpoint becomes a step in a flow and a tool for an agent.
Loom by Rutron
One process. One day. A running agent.
Loom is the platform your agents run on. Connect your APIs, draw the workflow, give an agent its tools, put a person at the decision points, and follow every run. One process becomes a running agent in a day.
In thirty seconds.
A working agent in one day
The building blocks are already there. One day is enough to connect your system, build the workflow and watch the first real runs.
A person at every decision point
Approval is a step in the workflow, not a setting. The run waits for the decision and records who took it.
What is inside.
01
Connectors
Import an OpenAPI or Power Platform definition, or describe an endpoint by hand. API keys, bearer, basic and OAuth2. Every endpoint becomes a step in a workflow and a tool for an agent.
- Import from OpenAPI, Swagger and Power Platform, with re-sync when the API changes
- API key, bearer, basic and OAuth2
- Secrets stay on the server. Test a connection before you use it
02
Workflows
Draw the process, or describe it in a sentence and let Loom draft it for you to review. Start it from a webhook, a schedule, a poll or by hand. Branch on conditions and test every step with real data.
- Describe a flow and review the draft before you keep it
- Triggers by webhook, schedule, polling or by hand
- If branches, typed forms, test a step with real data
03
Agents
Give an agent its instructions, a model of your choice, the tools it may use and a fixed output the next step can rely on. Optional access to your knowledge base.
- Instructions, a model of your choice, the tools it may use
- Fixed output fields the next step can rely on
- Optional access to your knowledge base
04
Approvals
A pause in the workflow where a person reviews, edits and approves or rejects, with an optional time limit. The run waits as long as it needs to and continues where it stopped.
- A durable pause, with an optional timeout
- Review the value, edit it, approve or reject with a reason
- The run continues exactly where it stopped
05
Runs
An inbox that puts “needs your approval” first. For every run you see the graph, the timeline, the agent’s conversation and every tool call.
- “Needs your approval” first, then every run
- Graph, timeline, the agent’s conversation and every tool call
- A run that got stuck never looks like a running one
06
Knowledge and LoomPilot
A knowledge base your agents can search, a playground to test what they will find, and LoomPilot, the assistant that answers questions about your workflows and runs and pins the answer as a live chart on a page.
- Documents your agents can search, with a test playground
- LoomPilot answers questions about flows and runs
- Pin the answer as a live card on a page
The agent day
One process. One day. A running agent.
Bring one process and access to the system it lives in. By the end of the day an agent runs it on Loom, from a simple check to a process with several steps, API calls, AI decisions and a person at the decision points. You watch its first real runs before you go home.
| When | What we do together | What you have |
|---|---|---|
| Beforehand, 30 minutes | Pick the process, agree who decides, arrange access to the API | A one-page brief |
| Morning | Map the steps and the exceptions. Import the API into Loom and connect it | Your system connected to Loom |
| Afternoon | Build the workflow. Give the agent its instructions and tools. Add the approval steps. Test with real data | A running agent with a person approving each action |
| End of the day | Watch the first real runs together. Decide what to fine-tune | The agent in your process, every run on record |
| Afterwards | Fine-tune. Decide what may run without approval. Manage it yourself, or we do | An agent in production |
One process. A system with an API or an OpenAPI description. Access arranged before the day.
Two ways to use Loom.
Self-serve
Your own login. Your flows, connectors, runs and knowledge in a UI built for operators. Change a step, add a connector, approve a run.
Fully managed
Our team builds the integrations, develops the agents and manages them in production. You review runs and approve where you chose to.
Built for operators.
Truth over polish
A failed or stuck run never looks fine. Status is shown as it is, with a symbol and a label, never by colour alone.
Unblocking is the job
When a run needs a person, that action dominates the screen. Everything else is context around it.
One vocabulary, everywhere
Status colours, badges and controls match across the inbox, the run detail and the builder.
How it compares.
Compared with generic automation tools
Approvals, a record of every run, agent tools and knowledge in one place. Connectors as settings, not code. A deliberately simple template language rather than a script engine inside your process.
Compared with chat assistants
An assistant gives answers. A Loom agent acts in your systems, with a person approving the actions you chose to gate.
Compared with a project plan
One day, not a quarter. The first agent runs before the first progress meeting would have been scheduled.
Questions about Loom.
What can be built in a day?
From a simple check to a multi-step process with API connections, AI decisions and human approval. Loom’s building blocks are what make a complicated process possible in a day.
Do we need Loom to work with Rutron?
No. Loom is how we build fast and keep people in control. If your situation calls for custom software, our AI software factory builds it.
Where does Loom run?
As a service in Rutron’s private cloud. You get your own login.
Can we change things ourselves?
Yes. Flows, connectors, approvals and knowledge are all in the UI. Or leave it to us and we manage everything.
Which AI models does Loom use?
Any model through your own key. Claude by default.
Which systems can it connect to?
Anything with an API. Import an OpenAPI or Power Platform definition, or describe the endpoints by hand.
What happens after the agent day?
You fine-tune the workflow, decide what may run without approval and switch it on. Yourself, or fully managed by our team.
Bring one process. Leave with a running agent.
Tell us the task your team repeats, the system it lives in and who decides. We come back within one working day with a date.
Plan your agent day