close fullscreen

External Agent Workflows+uAgents Agent Framework

help edit space_dashboard
Draft — This content has not been approved for publication.
ai generated
more_horiz
open_in_full page
fullscreen modal
edit edit
space_dashboard advanced

Verdict: [FETCHAI-AGENT-FRAMEWORKS-WITH-NARROW-HYPERON-BRIDGE-EXAMPLES]

uAgents (fetchai/uAgents, Python) is Fetch.ai's framework for autonomous, decentralized AI agents — the agent-network backbone of the ASI Alliance (the SingularityNET / Fetch.ai / Ocean merger that also stewards Hyperon). It is external augmentation technology: the framework core is non-Hyperon, but it is the decentralized transport/execution surface that lets Alliance members expose cognitive runtimes (including Hyperon MeTTa) to the network. Adopt the workflow — decentralized agent discovery + portable skills + the agent-fronts-a-cognitive-service pattern.

This is a sibling of the HERMES Agent Workflow — both are non-Hyperon-origin agent workflows to absorb. Where they overlap (the SKILL.md skill standard; MCP tool boundaries) is called out below.

The absorb-this workflow

  1. Decorator agent model. An Agent is an async runtime with decorator handlers (on_message / on_interval / on_event) and a Context (send/receive, per-agent storage, logging). Message exchange is envelope-based and cryptographically signed.
  2. Typed protocols + dialogues. Interaction is via typed Protocols (message-model digests, allowed replies, manifests); an experimental dialogue layer adds graph/state-machine structure (Node/Edge). Capabilities are first-class and discoverable.
  3. Decentralized Almanac discovery. The USP: agents register their cryptographic identity + endpoint metadata on the Almanac, a CosmWasm smart contract on the Fetch.ai ledger, and resolve peers by address — serverless, broker-free P2P routing. Registration is API-preferred with on-chain ledger fallback. (This is the only blockchain touchpoint — narrow registration/discovery, not a cognitive chain layer.)
  4. Crypto identity + mailbox transport. Local-wallet/Identity keys; an Agentverse-mediated mailbox lets NAT'd agents receive messages via a challenge/sign handshake.
  5. Portable SKILL.md skills. agentverse-skills is the first SKILL.md-format skill collection for the Fetch.ai/ASI ecosystem — on the same anthropics/skill-md-spec open standard as HERMES. Eight runnable skills (chat / deploy / image-gen / inspect / manage / memory / search / asi1-chat) that any AI coding agent (Claude Code, Codex, Cursor, Gemini CLI) can read and execute, driving the Agentverse via a relay that spins up a temporary hosted agent. agentverse-memory is itself an Agentverse Memory MCP service (episodic / graph / procedural / working memory).
  6. Adapters: MCP / A2A / LangChain / CrewAI. The uagents-adapter ships a code-backed MCPServerAdapter (ListTools/CallTool) plus A2A inbound/outbound and LangChain/CrewAI bridges. An experimental ChatAgent runs ASI:One (asi1-mini via litellm) with tool use. This places uAgents on the MCP-integration thread running HERMES (SC-3) → iCog mcp-xp (SC-6) → uAgents (SC-2).

The Hyperon bridge (narrow, example-scoped)

uAgents core is zero-Hyperon, but Fetch.ai's innovation-lab-examples ships three official examples where a uAgent fronts an in-process Hyperon MeTTa knowledge space — the genuine "agents-as-Hyperon-consumers" pattern:

  • web3/singularity-net-metta — a Medical RAG uAgent; web3/singularity-net-metta-fetch.ai — a Fetch/uAgents FAQ RAG; web3/singularity-net-metta-fetch-financial-advisor-agent — an Investment RAG.
  • Each imports hyperon.MeTTa (hyperon>=0.2.6), constructs an in-process MeTTa(), populates a native MeTTa space with structured atoms (metta.space().add_atom(E(S(…))) — not a vector store), queries via metta.run(), and returns results through the uAgents chat protocol with ASI:One as the NL layer.

Pattern to absorb: expose a Hyperon MeTTa cognitive service as a uAgent on the ASI-Alliance network — LLM for natural language, MeTTa for structured symbolic retrieval/reasoning, Almanac for decentralized discovery. Boundary: the coupling is example-scoped (confined to those three directories; the A2A/MCP examples are zero-Hyperon); these are Hyperon MeTTa-space demos — no AtomSpace / PLN / MORK / PeTTa.

Comparative placement

  • vs the Claw family (SC-7, MeTTa-native): the Claw runtimes reason inside the Space; a uAgent↔MeTTa bridge calls MeTTa as a library from an external Python agent loop.
  • vs metta-motto (SC-5): metta-motto calls LLMs from MeTTa; the uAgent bridge calls MeTTa from an LLM-fronted Python agent.
  • vs HERMES (SC-3): both are external agent workflows to absorb; HERMES is a single self-improving agent, uAgents is a decentralized agent network. They share the SKILL.md standard and the MCP tool boundary.

Provenance

Cross-Org Cognitive Repo Sweeps, SC-2 (Agent Frameworks / fetchai). S1 pins uAgents@c210bbea + agentverse-skills@16fab531; S2 pin innovation-lab-examples@71f35a5b. Three-model extraction (Codex citation-grade + Gemini synthesis + orchestrator spot-verification of the zero-Hyperon framework boundary and the in-process MeTTa bridge at the pins), 2026-06-24. Boundary: Fetch.ai blockchain/ledger/Cosmos/wallet/token infrastructure (cosmpy, fetchd, agent-launch-toolkit tokenization) is excluded as non-cognitive plumbing.

Verdict tags: [FETCHAI-AGENT-FRAMEWORKS-WITH-NARROW-HYPERON-BRIDGE-EXAMPLES] · [DECENTRALIZED-UAGENTS-FRAMEWORK-WITH-ALMANAC-BOUNDARY] · [PORTABLE-SKILLMD-AGENTVERSE-WORKFLOW] · [UAGENTS-HYPERON-METTA-BRIDGE-EXAMPLES-PLUS-A2A-MCP-INTEGRATION]