Put what your AI can see to work.

Workflows and agents act on the data altspark keeps current, so the recurring job reaches your team prepared for a decision.

  • Approval steps wherever a person should decide
  • Never writes into your database
  • No license for the people who use a Surface
Book a demo

Rebuilt by hand every week, the numbers arrive too late.

  • The recurring job, done by hand

    Invoices keyed, statements reconciled, the same report rebuilt every week.

  • Finding out too late

    The number that mattered was in last week's export, and nobody opened it in time.

  • Only people with a login see the number

    Clients, partners and most of the team have no seat in the source system, so they ask someone who does.

Put the data altspark keeps current to work.

  • Workflows and agents

    The work arrives prepared, and a person still decides.

    Steps read, draft and write back to apps with an API, and whoever builds the workflow puts an approval step wherever a person should decide. A run can pause for an approval, a question or an emailed reply.

  • Insights and alerts

    The number reaches you before you ask.

    A saved question is re-asked on a schedule and starts a workflow when the answer changes.

  • Surfaces

    Apps on your data, for people with no login.

    Full working apps on your data: people add and edit records in your stores, see live answers to custom queries, and view charts and maps. Private, public, password-protected or for a list of emails, and nobody who uses one needs a license.

Hand the gathering to altspark and keep the decision.

  • The weekly report, ready for review

    The weekly report and the month-end reconciliation are prepared from data kept current, and your team starts at the review.

  • Built by describing the job

    Describe the job in Claude, and the altspark MCP builds, runs and fixes the workflow with you.

  • Spend you can cap

    A run can carry a hard AI spending limit, drawn from the AI wallet or your own model key.

  • A check at every step

    Every step of a run validates its output against a schema.

  • Runs that start themselves

    A workflow starts on your schedule, on a webhook, or when mail reaches its own address.

altspark prepares, a person approves, your system posts.

altspark never posts into an ERP on your own server: the ERP validates and posts the file itself.

  1. Read

    The document and the system's records, read-only.

  2. Check

    Each document against them. Doubts stop and ask.

  3. Prepare

    An import file the system validates itself.

  4. Approve

    A person reviews it, at the approval step.

  5. Post

    Your system of record posts the file itself.

What customers say

Before altspark, I could build useful tools fast with Claude, but I had nowhere reliable to run them. Every tool meant another local process to babysit.
Gavin GuestVP of Product, Rosnet
Portrait of Gavin Guest

Security

You choose where a person approves.

  • A person approves wherever you choose.
  • Writes reach apps, never an external database.
  • Role-based access in every Surface.
  • Credentials encrypted, every action logged.
  • SOC 2 Type 1 report.

Pick the job your team does by hand every week.

altspark prepares it. Your team approves it.

Act on the tools you already run.

Sales & marketing

HubSpot, Salesforce, Pipedrive, Apollo, Clay and 13 more.

View all

Questions we hear

Only as you build it. A workflow runs on the trigger you choose, and an approval gate pauses it until the people you name approve or decline, in the app or straight from the email. Live insights, altspark's alerts, watch your data and, when a result changes, email your project team or start a workflow.

Yes, through steps you build. A workflow can call any app's API, and you can place an approval gate before the change goes out. altspark never writes to databases. For an ERP on your own server, altspark prepares an import file for a person to approve and the ERP imports it.

Two kinds. For your own AI steps and agents, you choose the model for each one from OpenAI, Anthropic or Google, and every plan can use your own API key for any of them. altspark's built-in AI, which researches and builds connections, helps build workflows and MCP servers, and writes Surfaces, uses the models altspark picks for each job, and it's included in your plan.

Each plan includes AI wallet credit every month: $25 on Spark, $50 on Ignite and $150 on Surge. When you need more, add funds to the wallet or turn on auto-recharge. AI usage is billed at the AI provider's price plus a small margin. Calls made with your own API key are billed by that provider and never touch the wallet. An agent run can carry a hard dollar cap, and AI calls stop when the wallet is empty, so there are no surprise charges.

Yes. The roles set on a data store follow its data everywhere: into Surfaces, MCP servers, AI tools and the run results people at the Use level see. A Surface or MCP server lets in only the roles you pick, and a field hidden from a role stays hidden there too.

Yes. Add their email addresses or their company's domain to a role, and let that role into the Surface. They confirm their email with a six-digit code, no altspark account needed, and see only the data and fields their role allows. A Surface can also be public, by link or with a password, when every data store it reads is open to the public.

Still doing the weekly job by hand? Let altspark prepare it.

Book a demo and bring the report your team rebuilds every Monday.

Book a demo