One data layer under the whole go-to-market motion: company and person data, waterfall enrichment, signals and intent, scoring and verification, lists, audiences and personalization, reachable from the console, the REST API, the MCP server and the CLI, with every write approved. The infrastructure under GTM, run by autonomous AI agents.
Company, person, enrichment, signals, intent, scoring, verification, lists, audiences and personalization behind one API and one console.
The console for the team, the REST API for your code, the MCP server for your agents and the CLI for your terminal, sharing one workspace and one ledger.
CRMs, sequence tools, ad platforms, sheets, docs and chat, each reached through an approval gate.
Every credit, every call and every approval in one audit trail, with rates shown before any work runs.
A company database, a contact tool, a verifier, an intent platform, a technographics feed and an audience uploader, each with its own credits, exports and column mappings. Most of the team’s data work is moving rows between them.
Sales scores accounts one way, marketing builds audiences another, RevOps cleans the CRM a third. The same company is a Tier A, a lookalike and a duplicate depending on who is looking.
An internal AI agent is only as good as the data behind its tools. Stitching six vendor APIs together, with six auth flows and six rate limits, is the project nobody finishes.
$ astrofabric run "ICP accounts hiring RevOps, buyer verified, push to HubSpot" → mission m_8f2 · 184 accounts · 271 verified · 92 staged $ astrofabric lists export l_41 --format csv → 92 rows · fit, signal, email_status, source $ astrofabric approvals approve ap_19 → 92 contacts upserted · HubSpot
Firmographics, technographics, funding, news, relationships, people, emails and phones, each field filled through a waterfall of licensed sources with a source and date, on demand or on a schedule.
Hiring, funding, technology change, news, relationships, growth, leadership and buying intent on the same canonical account, watchable as standing feeds with digests.
ICP scoring with reasons, bulk verification and hygiene, persistent lists that refresh, matched audiences and formatted exports, and drafts written from evidence, all from the same rows.
Every capability is a tool on the REST API, the MCP server and the CLI as well as the console, so your team, your code and your agents work from one workspace, one ICP and one ledger.
The ICP drafted from your site and confirmed, the suppression list loaded, the CRM and delivery tools connected, the credit ceiling and approvers set.
A plain-language objective in the console, a REST call from your code, a tool call from your agent over MCP or a command in the CLI. Same workspace either way.
Autonomous AI agents choose the capabilities the objective needs, run them in parallel against licensed sources and reconcile the results on canonical domain.
Contacts verified before they count, fit scored with reasons, duplicates merged and suppression applied on every list.
CRM upserts, sequence loads, audience pushes and exports staged as previewed and released by a person; reads and research never wait.
Schedules, watches and digests keep lists, audiences and the CRM current, with every run in the transcript and every credit in the ledger.
A signal-qualified, scored list pushed to HubSpot as opportunities, loaded into Lemlist paused, pushed to LinkedIn as a company audience and exported for Customer Match, from the same rows on the same morning.
An internal outbound agent connects to the MCP server and asks for the twenty best-fit accounts hiring RevOps with verified buyers, receiving JSON with signals, scores and sources instead of six vendor calls.
A weekly script enriches the active CRM segment, verifies contacts, builds the diff and opens the approval, with the ledger and transcript available for the audit.
The whole ICP watched for every signal family, a Monday digest to Slack ranking the accounts that moved, and the batch of new opportunities staged for one approval.
| By hand | With AstroFabric | |
|---|---|---|
| Vendors | Six subscriptions, six exports | One layer over a waterfall of licensed sources |
| Truth | A version per team | One ICP, one canonical account, one score |
| Access | Seats in each tool | Console, REST, MCP and CLI on one workspace |
| For agents | Six APIs to stitch | Every capability a tool over MCP |
| Governance | Whoever had the login | Approval gates, protected fields, one ledger |
| Cost | Paid for rows found or not | Rates shown first; misses on contact finding free |
An outbound or research agent you run connects over MCP and calls discovery, enrichment, verification and lists as tools, with the same approvals and ledger as the console.
A scheduled script on the REST API or CLI that qualifies accounts, verifies buyers, scores fit and opens a CRM approval every Monday.
Lists joined, deduplicated and charted in the code sandbox, results saved as lists and delivered to a sheet or Slack, on demand or on a schedule.
Three of a hundred. The full playbook library ships in every workspace, and anything you can describe becomes a mission.
Every capability that reads or researches runs inside the workspace credit ceiling without a gate. Every capability that writes to a CRM, a sequence tool or an ad account is staged for a person, from every surface.
Credits, calls, approvals and API keys are recorded per workspace, exportable, with rates shown before any work runs and a ceiling that stops it.
API keys carry scopes, connections carry the objects and fields they may touch, and approvers are named per workspace.
All data comes from licensed providers; the agents never scrape member networks, never send messages and never run campaigns in your name.
One governed layer under every go-to-market motion: company and person data, waterfall enrichment, signals and intent, scoring and verification, lists, audiences and personalization, run by autonomous AI agents and reachable from the console, the REST API, the MCP server and the CLI, with every write approved and every credit in one ledger.
Sales, marketing and RevOps teams that want one data layer instead of a stack of subscriptions, and developers or AI teams that want that layer as tools their own agents can call. The same workspace, ICP and ledger serve both.
Connect to the MCP server with a workspace key and every capability appears as a tool: discovery, enrichment, verification, scoring, lists, audiences and drafts. Writes still stage for approval, and the agent receives the same JSON, sources and dates the console shows.
From a waterfall of licensed sources for company, people, technology, funding, news and intent data, with each field carrying its source and retrieval date. No member networks are scraped, sources that disagree on a field are reconciled before the value is shown, and a contact lookup that finds nothing costs nothing.
HubSpot, Pipedrive, Attio and Zoho CRM; Lemlist and Instantly as paused sequences; LinkedIn Ads and Meta Ads as matched audiences; Google Customer Match and hashed CSV exports; Google Sheets, Docs, Notion and Slack. Each delivery passes an approval gate, and the API, MCP server and CLI use the same connectors.
No. Drafts load paused, audiences are created and opportunities are staged; sending, creative, budgets and campaign settings stay with your team in the tools they already use. The boundary is the same on every surface, so an agent calling the MCP server can prepare the work and cannot take the action that reaches a buyer.
Plans start at $49 per month and include credits. A verified contact is a few credits, company enrichment is a few credits per row, and scoring, lists, schedules, approvals and exports are free. Every rate shows before work runs and misses on contact finding are free.
Yes. Onboarding drafts an ICP from your website and the first mission runs against licensed data with results in the console and as CSV. Connect the CRM, sequence tools and ad accounts whenever you want delivery to go further. Nothing is written anywhere until a connector exists and a person approves.
One workspace, one ICP, one ledger. Ask from the console or call from your code, and the first governed list is ready in minutes.