Injective demos AI trading stack

Injective published the MCP server code as an open‑source repository under InjectiveLabs/mcp-server on GitHub. (github.com/InjectiveLabs/mcp-server) The MCP server exposes primitives for perpetual futures, spot transfers, cross‑chain bridging and raw EVM transactions that agents can call via the Injective SDK. (docs.injective.network/developers-ai/mcp) Injective’s release advertises a packaged developer surface of 22 tools across six categories and a test suite counted at 262 automated tests in the codebase. (ainvest.com/news/inj-launches-open-source-injective-mcp-server-enable-ai-trading-derivatives-markets-2603/) Private keys used by the MCP server are encrypted at rest with AES‑256‑GCM and scrypt key‑derivation, and transactions are intended to be signed client‑side through the Injective SDK rather than by a custodian. (github.com/InjectiveLabs/mcp-server/security) The Injective blog and docs note an Anthropic/Claude integration so MCP‑compatible agents such as Claude Desktop or Claude Code can query markets, open positions, monitor P&L and trigger signed on‑chain transactions via natural‑language prompts. (injective.com/learn, github.com/InjectiveLabs/mcp-server) Injective’s agent‑skills packaging lists seven modular trading skills — account management, AuthZ autonomous signing, deBridge DLN and Peggy cross‑chain bridging, chain analysis, real‑time market data, staking, and token metadata — and the skills are published in the Injective agent‑skills repo and discoverable via skills registries. (github.com/InjectiveLabs/agent-skills, lobehub.com/skills/injectivelabs-agent-skills-injective-trading-bridge) The MCP server’s tooling explicitly includes bridge support for Arbitrum, Base, Ethereum, Polygon, BSC, Avalanche and Optimism in its bridge skill documentation, enabling agents to route liquidity between Injective and major EVM chains. (lobehub.com/skills/injectivelabs-agent-skills-injective-trading-bridge)