Back to directory

Elemm

A dynamic, manifest-driven communication framework, the Landmark Manifest Protocol, that lets an autonomous LLM agent discover and call distributed APIs on the fly instead of relying on static, hand-written tool definitions.

Elemm

Elemm replaces static tool definitions with a manifest-driven architecture: instead of hard-coding every API an agent can call, an MCP gateway exposes nine core discovery tools, and the agent finds, navigates, and calls domain-specific actions dynamically as it needs them, for example being told to connect to an OpenAPI specification and list its available resources. Any Python function can be turned into a landmark that the gateway exposes, either as a FastAPI web service over HTTP and server-sent events, or as a standalone local MCP server over stdio, with schemas generated automatically from Pydantic models.

An interactive configuration generator detects the user's operating system and paths to produce a ready-to-paste client configuration for tools such as Cursor or Claude Desktop, and an optional dashboard shows token usage, manages stored credentials, and configures security policies. Built-in response controls limit how much data a single call returns to avoid overflowing an agent's context, sessions can be isolated to run parallel tasks safely, and a self-healing component gives the agent actionable fixes when a call fails.

More in Agent frameworks

See all 902
  • Recurse

    recurse.run

    A serverless harness that lets a coding agent design, test, and deploy its own specialist sub-agents as tools, MCP servers, or bots, defined by a schema-validated manifest.

  • ProofForge

    shulgin.is-a.dev

    An AI-agent pipeline that decomposes a mathematics problem, proves each piece, and formalizes the result in Lean 4 and Mathlib until the Lean kernel accepts it, with six of its proofs already merged into Google DeepMind's formal-conjectures repository.

  • local-coder

    github.com

    Builds a local coding-agent team on top of Ollama and OpenCode by detecting the machine's hardware, recommending models for seven specialist roles, and testing whether each assigned model can actually use tools, edit a repository and pass verification.