All docs

Docs · Get started

Thanks, Computer

Computational continuity for your work.

Most software forgets.

A ticket doesn’t know the objective it serves. An AI agent doesn’t know why it’s acting. An approval request arrives stripped of the context that created it. The work moves; the why stays behind.

Thanks, Computer (TxCo) is built the opposite way, believing that context, memory, and intent should travel with the work. Events arrive from anywhere — a web request, an inbound email, a cron tick, an AI agent. Every rule that resonates fires: invoking your services, running code, asking an AI model, or pausing while a manager weighs in. The results merge into one shared document, and that document — the what, still carrying its why — is the answer.

That continuity holds across every timescale of your work:

LevelWhat it isLives for
SagaA mission — why this mattersquarters, years
ArcA matter — what we’re resolvingdays, weeks
EventA beat — now, one thing happeningmilliseconds–days

Because every operation reads the same shared document, a participant three steps downstream — human or AI — still knows what matters and why. Use TxCo to run support, approvals, grants, or reporting — each as a small stack of plain-text rules that call your APIs or AI to move your work forward.

Underneath is a simple view: operational software isn’t an application, it’s coordination — between systems, services, people, and time. Today that coordination hides in glue code, queues, and inboxes. TxCo makes it a first-class, inspectable artifact: a flow you can read, version, trace, and share.

How it works

Work is broken into small steps called operations, organized into an op stack. Each operation is gated by a resonator — its firing condition, a few lines of TXCL: like a tuning fork, it rings only when the right kind of event passes by, and says what to do when it does:

WHEN @web.req.url.path == "/invoice"
EXEC "op://extract-invoice"
EMIT .processed_at = &now("rfc3339")

Everything at the same step runs in parallel; each operation’s output deep-merges into the shared document, which carries the flow to the next step. Operations execute only when their conditions match — nothing runs that doesn’t need to. An operation can be no code at all (the rule itself shapes the flow), any HTTP service in any language, sandboxed JavaScript running on the chassis, or an external AI tool. A stack of those steps is the whole department:

invoices
input100extract200enrichaudit300archiveoutput

Because the event is shared external memory — not parameters threaded through call chains — humans and AI work the same matter together: an AI op drafts, a human op approves, a service op ships, all reading and writing the same document.

What makes it different

  • Every protocol, one flow. Web, email, cron, TCP, and MCP ingress all land in the same envelope — the JSON document that carries the event and its context — and the same rules. The same stack that answers https://ops.example.com answers support@ops.example.com and answers an AI agent’s tool call.
  • Decisions are visible. Logic lives in readable text files, not YAML sidecars or buried application code. Every flow leaves a full trace you can replay step by step — debuggable for computers and AI alike.
  • The why travels with the work. Stamp an arc with the saga it serves and the objective rides in the envelope — services, AI, and humans all decide in context, instead of working from a goal they never met.
  • Built for waiting. Continuations let an operation suspend — for a slow model, a webhook, a human reviewer — and resume days later, exactly once, surviving restarts.
  • Nothing to deploy. One static binary is the whole chassis. Small logic runs as sandboxed Wasm on the chassis itself: no containers, no cold starts, safe multi-tenant isolation by default.
  • Any language, plain JSON. An operation is any HTTP service that reads JSON and returns JSON — if you can write a handler, you can write an op.

Where it’s going

Op stacks are made to travel. A stack is a versioned, signed artifact — distributable through standard OCI registries — so a working department (invoicing, triage, onboarding) becomes something you install, not rebuild. A growing fleet runtime adds custom domains, instant rollback, and a control plane, so ops.yourcompany.com becomes programmable operational authority. Delegate the subdomain and its DNS is handled — mail records, reputation keys, TLS — with every channel into your business (mail, web, agents) arriving at rules you wrote, can read, and can trust.

One platform, two ways to run it. Self-host the chassis — open source (Mozilla Public License 2.0), complete on its own, on your infrastructure. Or let our cloud run the fleet for you. Same stacks, same rules, nothing forked — txco apply targets either.

Try it in two minutes: Quickstart · txco demo

Edit this page · View as markdown