SOLAI Agent

🚀 SOLAI is currently in the Funding Phase on Pump.fun. Everyone can participate and become part of the AI + DePIN ecosystem from the beginning. 🔥 Buy $SOLAI: https://pump.fun/coin/Hy9XZ4Ae4oKtXYfuFzWkoNV18teCTpvWWu5PFD9Bpump

HOLDERS of $SOLAI will be able to rent their inference power to other SOLAI Coder users and earn $SOLAI in return.

SOLAI Agent is a local-first AI agent and coder built around the solai CLI.

SOLAI Protocol

The SOLAI protocol is under active development in this repository. It connects local agents, model providers, pricing, scheduling, signed heartbeats, telemetry, and provider discovery into a compute network for AI workloads.

The current protocol work includes:

• Provider identity with signed heartbeat payloads
• Local compute provider mode through solai provider
• Ollama-compatible inference routing
• Provider pricing per model in SOLAI/hour
• Availability scheduling for provider machines
• Runtime metrics for CPU, memory, GPU, model inventory, queue size and job state
• HTTP endpoints for health checks, metrics, heartbeats, jobs and chat completions
• A separate provider/ package so the compute layer can evolve alongside the CLI

Provider commands:

solai provider enable
solai provider status
solai provider price SOLAI-20B 4
solai provider schedule --from 22:00 --to 07:00
solai provider disable

The protocol code currently lives in this repository with the coder CLI because provider registration, local execution and agent workflows are being developed together. As the network layer expands, the protocol can be split into dedicated packages or repositories without changing the user-facing solai command flow.

Install

npm install -g @solaiecosystem/solai

Local model setup

Use the built-in provider flow to point SOLAI Agent at a local model server.

1. Open /provider_conf
2. Enter an address such as 127.0.0.1:11434
3. SOLAI Agent probes the endpoint, checks /api/tags and /v1/models, and stores the detected provider and model
4. Use /model to pick the model exposed by the server and adjust the reasoning effort for the current run

Useful commands:

/provider_conf
/model
/smartcontext true

For Ollama-backed models, /smartcontext true enables native smart context sizing. SOLAI Agent uses the active model context window to send the right num_ctx value to Ollama. Use /smartcontext false to disable it again, or set it directly in config.toml:

ollama_smart_context = true

Context control

SOLAI Agent keeps session history explicit and bounded.

/context
/smartcontext
/mini-model
/resume-type
/new
/clear
/compact

• /context sets the maximum context window before auto-compaction
• /smartcontext toggles Ollama native context sizing for local models
• /mini-model sets the model used for context compaction
• /resume-type sets the compaction strategy
• /new opens another chat during a conversation
• /clear starts a fresh chat
• /compact trims history when the conversation gets long

Monitor status line

Use /statusline2 to configure a second footer line with monitor data from the local model host. It can show CPU, RAM, swap, GPU, GPU model, VRAM, GPU temperature, Ollama runtime status, context usage, and GPU/CPU split.

/statusline2

You can also configure it in config.toml:

[tui]
status_line_2 = ["cpu", "ram", "gpu", "vram", "status", "context"]
status_line_2_use_colors = true

When a local provider is configured, SOLAI Agent reads monitor metrics from the provider host on port 9898 at /metrics; the line stays hidden until metrics are available.

What it includes

• Terminal workflow
• App server integration
• Python SDK
• TypeScript SDK
• Workspace-aware sandbox and approval rules

Support

SOLAI Agent is the free local agent and coder layer for the SOLAI ecosystem.

More

• Repository
• SOLAI website