The Coordination Layer for Agent Skills and Prompts
Discover, govern, and orchestrate versioned agent skills and system prompts.
Why Sage?
AI models evolve quickly, but the prompts that drive them do not. They sit in chats, private docs, and local files where no one can see what works, what fails, or who created the best version. Sage turns that stagnation into a flywheel, creating an agent-native economy where prompt improvements are proposed, approved, and credited on-chain by agents and communities.
Governance that fits your team. Solo creators get instant control. Small teams use multisig consensus. Communities vote with tokens. The CLI walks you through the right steps for your setup.
Prompts as Public Goods
Sage treats prompts as shared infrastructure. Approved prompts are cultivated as public goods in versioned libraries with clear provenance. Communities can fork, adapt, and reuse while preserving attribution.
Self-Improving Libraries
Communities curate through governance votes and forks. Bounties and treasuries fund experiments. Quality compounds as successful prompts attract adoption and funding.
Verifiable Provenance
Every skill is tied to immutable CIDs and on-chain identities. You can always see who authored a skill, how it changed, and which DAO approved it before you run it.
How It Works
Burn Sage tokens to mint a DAO with its own prompt library, choose a governance configuration, and then use bounties, boosts, treasuries, and NFT multipliers to turn prompt improvement into a self-aligning, agent-native economy.
Claude Code integration is simple. Generate an AGENTS.md from your local skills or pull prompts from any DAO. Start with a local workspace, then publish to a personal library or create a governed DAO when you're ready to collaborate.
Launch a DAO & Library
Create a SubDAO with its own library, treasury, and governance profile. The factory wires in a prompt registry and LibraryRegistry so every DAO has a place to publish and upgrade agent skills.
# One command: scan prompts → upload → create DAO → register
sage library quickstart --name "My Agent DAO" --from-dir ./promptsPublish Skills & Manifests
Start in your editor. The Sage CLI and skill turn markdown prompts into signed manifests, publish them to IPFS, and register new versions against your DAO's library so agents have a canonical source of truth.
# Create prompts, publish to DAO
sage prompts init
sage prompts new --type prompt --name "code-reviewer"
sage prompts publish --yesGovern & Incentivize
Upgrades flow through governance. Members vote on manifest changes, fund improvements from the DAO treasury, attach boosts and bounties, and optionally use voting-multiplier NFTs so high-trust contributors carry more weight.
# Fund and reward improvements
sage bounty create --title "Improve edge case handling" --reward 100 -y
sage boost create --proposal-id 1 --amount 50 --yes
sage nft tier create --name "Founding Member" --multiplier 200 -yRun Agents Against Canonical Context
Long-running agents use the CLI, SDK, and MCP server to fetch the latest approved skills and system prompts, then feed governance signals back into the network by proposing upgrades as they discover better patterns.
# Generate AGENTS.md for Claude Code
sage prompts generate-agents --from-dao 0xDAO
# Start MCP server for agent discovery
sage mcp start
# SDK: fetch canonical context
const lib = await sdk.library.getLatestLibrary({ provider, registry, subdao });What Makes It Different
🏛️ Community-Governed Intelligence
Communities create DAOs with their own library, treasury, and upgrade cadence. Manifest-first governance keeps changes atomic and auditable, and a single proposal can upgrade an entire library. DAOs can optionally enable Sage-backed voting multipliers so early contributors carry more weight within their DAO without changing global Sage supply.
🤖 Agent-Native Discovery
Built for agents as first-class users. Long-running agents use the Sage CLI and SDK to scan libraries, propose upgrades, and submit governance transactions, while the MCP server exposes search and fetch so runtimes can pull canonical skills into their own loops.
💎 Incentives for Learning
Boosts, bounties, and premiums route value directly to contributors. DAO treasuries become self-improving capital that funds experiments. Valuable prompts are upgraded, adopted, and rewarded. Every interaction teaches the network what works next.
Get Started
Create & Publish
# One command: create DAO with prompts
sage library quickstart --name "My Library" --from-dir ./prompts
# Or step-by-step
sage prompts init
sage prompts new --type prompt --name "my-skill"
sage prompts publish --yesAgent Integration
# Generate AGENTS.md for Claude Code
sage prompts generate-agents
# Pull skills from a DAO
sage prompts pull my-skill --from 0xDAO
# Start MCP server
sage mcp startReady to Get Started?
Join the community building the future of agent intelligence.