Your agents deserve better tools than a chat window.

No digital employees. No agentic revolution. Tools that return JSON and exit 0.

receipts ask "what is prospera" --json

It returns JSON. It exits 0. That’s the whole trick.

Self-describing offline. Budget-metered. Stable exit codes. Your agent reads the manual in one call. When did your SaaS last do that?

One JSON envelope for every response
Stable, documented exit codes
Self-describing offline — schema, flags, errors from the binary itself
Budget-metered — every call reports what it cost
The agent is the primary user; the human reviews

{
  "schema": "receipts.cli.response.v1",
  "ok": true,
  "command": "ask",
  "data": {
    "question": "what is prospera",
    "outcome": "answered",
    "claims": [
      {
        "claim": "Próspera is a ZEDE in Honduras.",
        "sourceUrl": "https://example.com/source",
        "verdict": "supported",
        "published": "2026-07-01"
      }
    ],
    "searchTrail": [{ "query": "prospera law", "results": 4 }],
    "uncertainties": []
  },
  "costDollars": { "model": 0.09, "search": 0.04, "total": 0.13 },
  "diagnostics": { "durationMs": 12100, "retries": 0 }
}

0ok

2auth

4network

6rate limit

10partial — budget hit, work reported

The kit

Six tools shipped, one loading. Zero dashboards. Nobody asks you to book a demo.

receipts

GitHub

RESEARCH THAT SHOWS ITS SOURCES. 14 SECONDS. FIFTEEN CENTS.

receipts ask "<question>" --jsonbrew install treygoff24/tap/receipts14.4 s · $0.15 · 19 supported claims (live)

lens

GitHub

FINDS THE PHOTO YOU MEANT. 1,100 IMAGES FOR $2.21.

lens find "<description>" --jsonbrew install treygoff24/tap/lens1,100/1,100 images · 98 s · $2.21 · search ~2 s

scout

GitHub

READS THE REPO SO YOUR FRONTIER MODEL DOESN’T HAVE TO. QUOTES, OR IT DOESN’T SHIP.

scout "<question>" <dir> --budget 8kbrew install treygoff24/tap/scout100% recall · 0 poison survivors · ~$0.10/query

exa-agent

GitHub

68 COMMANDS. DESCRIBES ITSELF. NO DOCS TAB.

exa-agent capabilities --compactbrew install treygoff24/tap/exa-agent68 commands · self-describing offline

elv

GitHub

320 VOICE OPS WITHOUT THE MCP CIRCUS.

elv capabilities --jsonnpm install -g eleven-agent-cli320 operations, one envelope

delegate

GitHub

SIX MODEL FAMILIES. ONE DISPATCHER. ZERO LOYALTY.

delegate <lane> work --prompt-file <task>pip install delegate-agent-cli6 child runtimes · safe/work isolation

law

SOON

WILL CITE REAL CASES. IMAGINE THAT. (LOADING.)

law research "<question>" --jsonin the chamber — not yet shipped

Receipts, or it didn’t happen

Measured on live runs. Screenshots available. Bring your stopwatch.

●14.4 s — receipts: source-verified research brief, 19 supported claims, $0.15●98 s — lens: 1,100/1,100 images indexed for $2.21; search in ~2 s●<$20 — inference cost to build each tool, one evening each●528 ms — ambient copilot tick — full context read + suggestion●14.4 s — receipts: source-verified research brief, 19 supported claims, $0.15●98 s — lens: 1,100/1,100 images indexed for $2.21; search in ~2 s●<$20 — inference cost to build each tool, one evening each●528 ms — ambient copilot tick — full context read + suggestion

14.4 sreceipts: source-verified research brief, 19 supported claims, $0.15
98 slens: 1,100/1,100 images indexed for $2.21; search in ~2 s
<$20inference cost to build each tool, one evening each
528 msambient copilot tick — full context read + suggestion

All figures from live acceptance runs, 2026-07-01. Local builds, real data, real APIs.

Process is a product

The method that built all of this builds your thing too. It was shipping autonomous loops in January 2025, before it was cool.

Fieldcraft started as a bet in January 2025: an agent loop that couldn't declare itself done until the work actually was. That became the autonomous loop — the /goal architecture, completion enforced by a Stop hook — shipped long before Anthropic or OpenAI offered anything like it natively.

The loop grew into an autonomous dev kit. It built delegate, which put every model family on tap. delegate grew into foundry: waves of different models implement, review, and fix each other's work while a coordinator re-verifies every claim. Then foundry started shipping products — receipts and lens were each built in one evening, for under $20 of inference, receipts attached.

The tools build the tools. This site was built the same way — and the log of what broke is in the repo.

Built by Trey Goff and the team behind Praxient.

wrightBuilds and maintains the agents.
memorumMemory that survives across sessions.
agentlintersLint rules for agent-written code.
llm-councilMultiple models deliberate hard calls.

Work with the team

Fieldcraft is the shop floor. Praxient is the room with the client. Same discipline, aimed at the operation. They wear the suit; we drive the truck.

Work with Praxient