Most AI tools give you an output, then discard the structure that produced it. Meterless preserves that value: Relay turns work across your applications into reusable missions, while Gaia keeps your memory, goals, decisions, and artifacts together so your AI work compounds across sessions, devices, and models.
AI models will keep changing. The lasting value is the structure created around your work: the missions, memory, decisions, workflow state, policies, artifacts, and proof of what succeeded. Meterless keeps those assets on your device and independent of any one provider. Relay makes successful workflows reusable. Gaia carries the context forward. Change models without rebuilding how your work gets done. Run once with a frontier model then reuse with cheaper models at scale.
Both sides do the job. One throws away the plan, the reasoning, and the memory when the session ends. The other saves them to your device: replay the saved structure for 0 new tokens, edit it, and let it get better every run.
Describe the goal, select the applications Relay can access, and inspect the plan before anything runs. Successful work becomes a mission you can verify, schedule, and replay, with vision-verified steps and approval gates.
Gaia keeps your projects, memory, missions, research, and artifacts connected across sessions, and independent of whichever model created them. Instead of starting from zero every time, you build on what has already been learned.
Explore Gaia →Convert files, edit documents, process images, and run coordinated AI agents directly in your browser. No account required. Your files are processed on your device, not uploaded to a processing service.
Type one goal and Swarms auto-generates the whole agent team, then runs it in parallel, locally in your browser. The graph is yours: edit it, rerun any branch, refine it over time, and keep every artifact with its lineage. Free and open source.
The harness, memory, workflows, and control layer remain under the user’s control while the underlying intelligence can come from local or cloud models. Relay, Gaia, and Swarms share the same Meterless primitives for context, state, orchestration, and verification.
See the work, not just the claim
8 windows
Relay · per mission
Missions drive up to eight real application windows with vision-verified checkpoints and approval gates you can replay.
300+ agents
Swarms · 64 in parallel
One goal fans out to a graph of specialized agents. Every run reports what was spent per accepted result; the graph and artifacts are yours to inspect.
12 brains
Gaia · shared memory
A twelve-brain planner over persistent, user-owned memory stored as readable files. Grep it, audit it, take it with you.
7.3–15×
Fewer tokens · the stack
The same work for a fraction of the tokens. Up to 90% less token spend. The Apache 2.0-licensed engines are on GitHub: read the source, run it yourself.
This engine isn’t public yet. Drop your work email and you’ll get the source link the day it ships. No spam, no marketing sequences.
We’ll send the Swarm source link to your inbox the day it ships.