Skip to main content

a9script connects your systems through their APIs and runs the automations between them to the end. Every call to another system is configured once, tested before it goes live and recorded as it runs. The code that remains is the business logic in plain JavaScript, written by your team or by an AI agent such as Claude Code. Integrations sets it up for your business, as a fixed-price project or by the hour.

a9script comes from Andreas Lengauer, a long-term partner of Integrations. Any endpoint can serve your APIs to AI assistants as an MCP server. The data sits on one server in Germany.

The a9script dashboard with sample numbers for arrivals against throughput, backlog, the week with an activation and the flow into scripts and out to connections

Key Features

One a9script endpoint with its four readings: webhook, API, MCP server and app

Endpoints and MCP Server

An endpoint set to MCP lists the calls you configured as tools for an AI assistant, each with a generated schema. The credential stays on the platform. MCP is the open standard AI apps use to work with tools and data. Through it an assistant can call a legacy REST API with no server written for it.

The same endpoint can be read as a webhook, which acknowledges the sender at once and runs your scripts afterwards, as an API, where the caller waits for one script's answer, or as an app that serves a folder of files with sign-in. Several scripts can listen on one webhook, each with its own filter. Every activated endpoint describes itself.

The a9script explorer beside a connection with its address, identity and endpoints

Connections and Secrets

Every system you connect has one place with its address and its credential. The platform keeps the credential away from every script, attaches it as the call leaves, refreshes an expired token once and sends the call again.

You enter a secret once. It stays encrypted at rest and appears on no screen, in no command and in no recorded request. Sandbox and production each hold their own connections and credentials. The same script runs in both, with no credential carried across by a promotion.

A configured call in a9script after Test, with the answer the script gets and the request sent, the credential named and never shown

Configured Calls

A call to another system is configured once: address, credential, retries, paging and what counts as an answer. You name it in your own words, and the script calls it by that name. Test shows the answer the script gets and the request that was sent, with the credential named and never shown.

The script editor completes every configured call with its signature and answer on hover, and marks a name that resolves to nothing while you type. Your team tests a call before it is activated.

Calls to HTTP APIs, mail, CSV files, mappings and AI models are configured the same way. A model call holds prompt, model and the connection with the key in one place, and the answer of a model is input the script validates.

A traced test run in a9script with every outbound call, what it sent and what came back

Workflow Runs with Retries

A run is written down step by step. A transient failure is repeated, three times by default. A rate limit parks the whole run for the time the other system asked, up to 24 hours, holding nothing open. After a restart, a deployment or a lost machine the run continues from its last saved point and makes no call twice.

In the sandbox you start a run with trace and read every call, what it sent and what came back, kept for 7 days. A stepped run parks after every external call and waits for your check before the next step.

Main Features

Pre-configured Webhook Senders

Senders such as Atlassian Jira, GitHub, Linear, Microsoft Graph, Okta, Shopify, Slack, Stripe, Twilio and Zoom come pre-configured, verification included. Handshakes are answered without a run. Any other sender becomes a custom recipe.

Two-Way Data Sync

Every record a script creates in another system is matched to its source id in a store per environment and updated by the next run. Ignore rules drop your own writes coming back before any script runs, which keeps both systems in step.

Thousands of Records a Night

Thousands of records a night are read page by page, mapped, checked for duplicates and written in bounded pieces. A bad row is set aside with its key and reason. The job goes on and, after a pause, resumes at its place.

Alerts That Stay Red

A failed run turns its script red until a person acknowledges it, and a connection that meets faults in a row opens its circuit and turns red as well. Notification rules mail you, throttled, and system triggers start a script of your own.

Versions with Diff and Restore

Every activation is a numbered version with history, diff, blame and restore. Production is marked in colour, and activating or deleting there asks first. A promotion from the sandbox shows a plan before anything changes.

Texts Owned by Your Team

A mail, a ticket comment or a CSV mapping is a template kept outside the code. The person who owns the wording, a support or operations engineer, changes it, tests it with a sample and activates it. The script keeps calling the same name.

  • Mail through SMTP or Microsoft Graph, with attachments
  • Mail sent, redirected to one inbox or recorded, per mail account
  • Runs started by a schedule or by a mail in a Microsoft 365 or IMAP mailbox
  • PDF documents built from blocks or markdown
  • Large files held by the platform, CSV of any size in chunks
  • GraphQL APIs called like any other HTTP API
  • A command line for building and running from a terminal
  • Your own users signed in at your endpoints, with your roles
  • Events no script takes, counted and logged
  • Security-scanned builds

Included Modules

The a9script Partners page with sample numbers for latency, repeats and an open circuit per connection

Insights

Every recorded number that informs a decision is drawn on a page that names its question and its decision. The pages hold counts, durations, codes and names, never the content of a message.

  • Dashboard with arrivals against throughput, backlog and failed runs
  • Errors by code, the code new since the last activation named first, to fix or roll back
  • Durations with working time against wait for a slot, to tell code from capacity
  • What comes next: the next 24 hours of scheduled starts and resumed runs
  • Partners with latency, calls, repeats and an open circuit per connection
  • Node load for the administrator: slot occupancy, memory and the system log
A configured function in the a9script editor with its signature shown on hover

Building with Agents

An AI agent builds integrations from a terminal or over MCP and never holds a credential.

  • MCP in both directions: build over MCP and serve over MCP, with the credential on the platform's side both times
  • An agent probes a call, writes a folder, applies it, runs it with a trace and reads what came back
  • A built-in guide teaches an agent the whole loop and is checked against the product on every build
  • Claude Code connects to your sandbox with a token for one environment that expires within 14 days
  • A Jira project connected to a Linear team by an agent, with the existing connections reused and what it could not verify reported
a9script with production marked in colour and the confirmation naming the environment

Security

Tenants and environments are sealed off from each other, with every security statement held true by the product's own tests.

  • Secrets encrypted at rest and returned by no screen, command or diff
  • Tokens that name one tenant and environment and expire after 14 days at most by default
  • Password plus second factor with recovery codes, and sessions that end after 30 minutes idle
  • Roles for Administrator, Tenant admin, Specialist and Auditor, and no default administrator
  • Network allow-lists for inbound calls and sender signature checks
  • Two append-only audit trails, hash-chained to show tampering
  • Backups taken while it runs, and a deleted tenant removed from every machine

Set Up for Your Business