ANCHOR 163 // home
// services

Systems that run themselves.

AI agents and automations that do a recurring job completely, built by a team that runs the same systems on itself first.

// on this page · 4 sections
scoped builds$15k–$40k
single automationFrom $5,000
managed from$400/mo
first conversationFree
Anchor 163 builds software and runs its own research and trading systems. We build AI agents, automations and data systems that run on their own, then either operate them for you or turn them over to your team.
The automations and page engines below run on our own products first, so what we build for you is what we already operate.

What we build

AI agents

Systems that read your email, documents and data feeds, then act on them by drafting, filing, flagging or answering.
Built on the Claude API, which we use in production daily.

Automation pipelines

The recurring work nobody should do by hand. Data gets pulled on a schedule, reports get generated and numbers get reconciled to source.

Custom software

Dashboards, internal tools and integrations.

SEO page engines

Sites that generate hundreds of local pages from data, one for each town you serve and each service you offer.

Any model, swappable

The model sits behind an interface, so it can be replaced without touching the system around it. We use Claude daily and reach for other models where they fit better.
Providers change pricing and capability monthly. A build hard-coded to one vendor is a liability you inherit.

Who we build for

Teams putting Claude to work

API integration, agent workflows and MCP, with an honest read on what should not be automated yet.

Trading firms and funds

Market data pipelines and execution tooling, with reconciliation that ties to the venue's own records.

Prediction markets

Execution on Kalshi and Polymarket, with order state that survives a disconnect and positions tied to the fills.

ML teams and federal programs

Independent checks on whether a model's numbers hold up, and open-source code for Flower and NVIDIA FLARE that catches a node the fleet average hides.

Accounting firms

Document intake with confidence scores and nightly reconciliation. An exception queue replaces data entry.

Local service businesses

The office work, the follow-up you never get to, and a web presence that maintains itself.

See the work

A website that updates itself

A tree service's site with 332 generated pages and a daily note written from the weather. It costs about $45 a month to run.

Three worked blueprints

Blueprints for composite businesses, each drawn by our blueprint desk with a cost band.

The Kalshi API in practice

What the docs say, and what happened when we ran automated systems against it with our own capital.

telltale on GitHub

Open-source code that flags a federated learning node training on bad data. Runs inside Flower and NVIDIA FLARE. MIT licensed.

How an engagement works

  1. Describe your business to the architect for a free draft, or bring the task straight to us. The first conversation is free, and you leave with an honest read on whether software should be doing it.
  2. We scope one build and deliver it. A single automation starts at $5,000, and a scoped build across several systems lands between $15,000 and $40,000.
  3. We watch a quiet job and fix breaks from $400 a month. We run it for you from $800 a month, or $1,500–$3,000 across several systems. Or your team runs it.
Anchor 163 LLC is also registered for federal work, with SAM.gov active and a CAGE code assigned. We take contract engagements, including independent AI model evaluation. Details are on the capability page.