The local-first context stack for agentic AI.

Agents fail on thin context, not weak models. Meterless is the context layer: memory, world modeling, reasoning compression, coordination, execution, and runtime quality, local-first.

The proof runs on your machine in 90 seconds, no API keys: the same 12-step task, cold vs warm. The warm run starts with H-MEM memory and finishes in 8 chunks instead of 12 (output below is real, tokens estimated at chars/4):

npx tsx examples/memory-compounding-research/index.ts

At 20 steps, the Markovian cost model saves 86% of input tokens vs naive history accumulation (modeled; the worked example is in engines/markovian/docs/efficiency-model.md). Measure it yourself: cd engines/markovian/reference && npx tsx scripts/measured-run.ts.

User / Event
   ↓
Scout Intent
   ↓
H-MEM + World Model
   ↓
Markovian Reasoning or Swarm Coordination
   ↓
Runtime Quality Layer
   ↓
Relay Execution / Gaia Interface / Swarms Output
   ↓
Verified Output + Updated Memory

One architecture. Three product surfaces prove it. Read the full flow in docs/architecture/stack-overview.md.

Implement an engine spec. The engines are AGENTS.md-driven implementation specs with runnable references and conformance suites. Clone one folder and hand it to your coding agent:

npx degit meterless/meterless/engines/hmem my-hmem
cd my-hmem/reference && npm install && npm test    # a working reference, green in about a minute
npx tsx ../examples/01-add-memory/index.ts         # watch a memory get mined, stored, audited

Your agent is done when the conformance suite passes: HMEM_IMPL=<your build> npx tsx conformance/runner.ts.

Start from a template. npx degit meterless/meterless/templates/agentathon-starter my-agent gives you a memory agent that runs in two minutes (templates/).

Read the architecture. Start at docs/architecture/why-meterless.md.

Get the products. Installers live on the app repos: Gaia, Relay, Swarms.

Roadmap: swarm-orchestration, runtime, and fulcrum drop as new engine folders on a monthly cadence. See ROADMAP.md. Unreleased engines have no folders here yet, by design.

Three product surfaces, one architecture. Binaries are proprietary and ship from the per-app repos. Full docs live here under docs/products/.

Gaia is the personal agent workspace. Docs · Install

Relay is the agent execution layer. Docs · Install

Swarms is the divergent generation layer. Docs · Install

This repo is built to be worked on by coding agents. Every engine is a spec a coding agent can implement, and a conformance suite that proves the agent got it right (engines/<name>/conformance/). AGENTS.md routes an agent to the one folder it needs. llms.txt gives non-coding agents the same map. Each engine folder is self-contained, with its own full-length AGENTS.md.

Engines drop monthly as tagged releases on this repo. Next: swarm-orchestration (August 2026), runtime (Q4 2026). Targets, not promises; dates move, the cadence does not. Details in ROADMAP.md.

Contributions are welcome. Start with CONTRIBUTING.md.

Everything in this repo is Apache 2.0. Product application binaries are proprietary and live in the per-app repos.