Skip to main content
Meteora ships an official Agent Skill — a portable instruction pack that teaches AI agents how to do things on Meteora correctly: launch tokens, create and seed pools, swap, manage positions, claim fees, and write integration code against the real SDK APIs. It lives in the meteora-invent repository and follows the open Agent Skills format, so the same folder works in Claude Code, OpenClaw, Hermes, Cursor, and any other Agent-Skills-compatible runtime.
The skill and the Documentation MCP are complementary. The skill is the operational playbook (which command, which config, which safety gate); the MCP and llms.txt are the always-current reference layer the skill tells agents to consult for depth.

​
What the skill teaches an agent

ACT: run studio actions

All 38 Meteora Invent studio actions with their real flags and config blocks — launches, migrations, pool creation, liquidity seeding, swaps, position and status reads, fee claims, alpha/presale vaults, locks, and Stake2Earn farms.

BUILD: write correct SDK code

Version-pinned reference packs for the DBC, DAMM v1, DAMM v2, and DLMM TypeScript SDKs — verified client setup, method maps, runnable flows, and “version fences” that call out breaking changes agents commonly get wrong.

Safety gates

Devnet-first, dry-run-first, explicit owner confirmation before mainnet execution, on-chain verification after every action, and key hygiene rules (an agent following the skill never asks you to paste a private key into chat).

Intake contracts

Per-flow checklists of exactly what the agent must collect from you before executing — amounts and units, addresses, curve parameters, funding — with documented defaults for everything else.

​
What’s inside

​
Install

Working inside a meteora-invent checkout? The skill is discovered automatically from skills/. To make it available in every project, copy it to your user-level skills directory:
Terminal

​
What you can ask for

Once installed, the agent routes plain-language requests to the right protocol, path, and commands:
  • “Launch my token on a bonding curve on devnet, graduating at a 600 SOL market cap”
  • “Swap 0.5 SOL into this DAMM v2 pool” or “buy 0.2 SOL of this DBC token”
  • “Show my DLMM positions on this pool and claim all my fees”
  • “What’s the graduation progress of my DBC token?”
  • “Migrate my graduated DBC pool to DAMM v2”
  • “Write a TypeScript bot that rebalances my DLMM position around the active bin”
The skill makes the agent gather required details first (network, wallet, amounts with units, addresses), simulate with dryRun before executing, ask for your explicit confirmation before any mainnet transaction, and verify the result on-chain before reporting success.

​
How it stays accurate

  • Version-pinned: every SDK reference pack states the exact package version it was verified against, with breaking-change “fences” for older APIs.
  • CI-validated: the meteora-invent CI checks the skill’s frontmatter against the Agent Skills spec, fails on broken internal links, fails when SDK version pins drift from studio/package.json, and keeps the bundled config templates byte-identical to studio/config.
  • Source-verified: action flags and config semantics are generated from the studio source, not written from memory.

skills/meteora on GitHub

Browse the skill source, reference packs, and config templates