Your recurring work, with your sign-off.
Any industry, one infrastructure. From social media to customer support, e-commerce to finance, research to software: you describe the work in the chat or build it in the workflow editor; agents run on your own machines connected to your own server, and you approve the critical steps. These are example setups; the screens are drawn with demo data.
Every area has its own page: chat, workflow, setup.
Each card is one team’s recurring work. Its page shows the chat screen, the workflow, the setup and how it connects to the tools you use.
- 01Content and marketingSocial mediaThe week’s posts, waiting for your approval on Monday morning.Open the example
- 02Content and marketingContent and SEOThe topic is yours; research, draft, SEO checks and a CMS draft come from the workflow.Open the example
- 03Sales and supportSalesA form submission lands in the CRM with its score and the reasons.Open the example
- 04Sales and supportCustomer supportTickets get sorted and replies drafted; you send them to the customer.Open the example
- 05Operations and financeE-commerceA hundred product descriptions overnight; only the ones you approve reach the shop.Open the example
- 06Operations and financeFinanceMonth-end reconciliation: the files are yours, the list of differences is on your desk in the morning.Open the example
- 07Operations and financeData collectionRegular data from permitted sources, in one table.Open the example
- 08Research and documentsResearchA sourced market report on your desk every Monday.Open the example
- 09Research and documentsContractsDrop in the contract pack; get a clause-by-clause checklist back.Open the example
- 10Software teamsSoftwareA shared queue, shared memory, clear roles.Open the example
Installed on your server, connected to your tools.
Remyral runs on your infrastructure: the management side on your Linux server, the work on your machines. You connect to the tools you use with workflows and automations: your tool’s API, signed webhooks and MCP. The key in the vault, the address on an allow-list, the critical step behind an approval.
Step 01 / 06
Installed on your server
Your IT team · with us during the pilot
Console, API, database, queue, vault, audit log and backups run with Docker on your own Linux server; the installer starts by verifying the release signature. The workers that run the jobs are your machines and connect outbound to the server; no inbound ports are opened. Work files, records and keys are stored on your infrastructure; content processed by AI goes from the worker to the model provider you choose, and data sent to external tools goes to the connections you set up.
Step 02 / 06
Seal the key in the vault
Your admin
Your tool’s API key is sealed in your browser and bound to the address you allow; the control plane cannot decrypt it. The AI agent never sees the key; only the HTTP step uses it. Your admin enters the value.
Step 03 / 06
Allow the address
Project admin · authenticator code
You add the address of the tool you connect to the project’s HTTP allow-list. Requests always leave from your worker; local network, private and metadata addresses are always blocked.
Step 04 / 06
Build the flow
Together, or your team
We don’t impose a ready-made flow: design yours with AI by describing it in the chat, or yourself in the workflow editor; you can also start from a template. An AI step (with an output schema), an approval and the HTTP step on one canvas; you set the HTTP step’s fields to match your tool.
Step 05 / 06
Approve what goes out
Authorised team member · project admin
A person reads and approves before anything is written outside. Saving a workflow with an HTTP, email or code step needs a project admin and a fresh authenticator code.
Step 06 / 06
Try, trigger, watch
Your team
First you run it by hand and see each step’s input and output; if something goes wrong, it says plainly why. Then a schedule or a signed webhook starts it; outgoing webhooks and email report back, and setting changes land on the audit log.
Ways to connect
- outboundWorkflow HTTP step
GET, POST, PUT, PATCH, DELETE; the credential comes from the vault, for allowed addresses.
- outboundEmail step
SMTP with TLS; the password in the vault.
- eventsOutgoing webhook
Job, approval, deploy, workflow and cost events, signed, to n8n, Zapier or your CI; delivery log and retries. No prompts, logs or secrets in the body.
- inboundSigned webhook
HMAC-SHA256 with a timestamp and a one-time value; optional IP list, body up to 256 KiB. Starts a workflow or a job.
- inboundAPI token and service account
Tokens limited by scope, lifetime and IP range, plus an OpenAPI document; another system opens jobs or asks for status.
- agentMCP server
Switched on per project; allow, ask or block per tool; every call on the audit log.
Good to know
- If the tool you use has an API, you connect with the workflow HTTP step; the ready templates come with a generic request and you set the fields to fit your own setup.
- Incoming webhooks are verified with Remyral’s signature. If your tool doesn’t produce it, you put a small signing step in between (n8n, Zapier, Make or your own function), or Remyral pulls from the tool’s API on a schedule.
- In the pilot we first define the need, install on your server, then build the flow together and test it with sample data before using real data.
True for every use case
- The installation is on your infrastructure: work files and records are stored there; content processed by AI goes to the model provider you choose, and data sent to external tools goes to the connections you set up.
- Password- and key-like values are masked; in flows with personal data you set the retention period and privacy notice to match your data protection policy.
- Outputs are delivered as CSV, Markdown, JSON, PDF and images; CSV opens straight in Excel.
- We recommend flows that leave the decision with a person: a team member reviews the outputs, and the critical step moves on with an approval.
Which one is closest to you?
Whatever your industry, one infrastructure: Remyral runs on your own server and we build your flow together in the pilot. Just describe the work you want to simplify and the tools you use in your application.