In recent years, multi-agent systems have shifted from rule-based, symbolic agents to sub-symbolic agents built on language models. However, several limitations remain unresolved. The new frontier is neuro-symbolic agents, which pair language models with deterministic tools and explicit protocols. Without a common framework to standardize and compose them, each study rebuilds the same infrastructure, and results are hard to replicate. This paper presents EREBOTS 4.0, an open-source, hybrid-AI, multi-agent framework that fills this gap for conversational assistants. Language-model agents plan and route requests, while deterministic tools compute exact results over the Model Context Protocol (MCP). An orchestrator selects one or more specialist agents per turn and calls them over the Agent2Agent protocol (A2A). A new agent, local or remote, is added by listing its URL in a configuration file. The framework also manages identity, user context with consent flags, task states and traces. Its A2A server passed 97 tracked requirements in the official A2A Technology Compatibility Kit, with no failures.
Figures & tables
Nr.
Code metadata description
Value
C1
Current code version
1.0.0
C2
Permanent GitHub link to code/repository used for this code version
Python ≥ 3.13 and an OpenAI-compatible LLM endpoint. A checkout resolves from a uv lockfile, and the published wheel bounds the three SDKs the framework codes against. Tool-using specialists need a model that supports tool calls. Docker Compose provides an optional deployment stack
Table 1: Code metadata
Figure 1: EREBOTS 4.0 architecture. One user turn: (1) message, (2) context loading, (3) LLM planning, (4) A2A delegation, (5) MCP tool calls, (6) streamed answer, (7) persistence. Dashed lines: possible process or host boundaries.
Evaluation unit
Passed
Failed
Skipped
Not tested
Tests, /a2a/tck
180
0
85
n/a
Tests, /a2a/tck-minimal
140
0
125
n/a
Requirements, /a2a/tck
94
0
10
25
Requirements, /a2a/tck-minimal
73
0
30
26
Requirements, union of both runs
97
0
7
25
Table 2: TCK counts for version 1.0.0, out of the 129 requirements the suite tracks. The two endpoints cover different requirements by construction. The minimal card declares three capabilities absent, which is why 30 requirements are skipped against it, and without streaming it offers no Server-Sent Events to exercise.
Figure 2: The Vue [ 31 ] client during the clarification scenario. The orchestrator routes to PersonalAgent , which asks for the missing value it does not have, and every turn carries its trace and its timings. The question comes back in Italian because the stored preference is already in the prompt.
Figure 3: The clarification scenario seen in PostgreSQL. The task sits in TASK_STATE_INPUT_REQUIRED while the question is open, completes once the user answers, and the stored profile changes from Italian to Spanish.
An interaction protocol formalizes how the agents in a multiagent system interact, which facilitates implementing agents. Existing approaches yield agent implementations specific to the selected protocols. How can we engineer intelligent agents that can enact protocols but are programming-free? Our contribution, Ahoy, addresses this question by creating LLM agents that dynamically select and enact declarative protocols to achieve user goals. We demonstrate that an Ahoy agent can correctly and intelligently enact multiple protocols - concurrently if appropriate to the user goal - without specialized training. Ahoy's significance lies in that it brings together declarative protocols and LLMs, both approaches that promise improved knowledge engineering for agents.
Omkar J. Joshi, Munindar P. Singh, Amit K. Chopra
North Carolina State University, Raleigh, NC, USA · Lancaster University, Lancaster, UK
When will robots have their ChatGPT moment? Such a breakthrough requires a general-purpose robot that can handle unfamiliar tasks in unfamiliar environments, remain controllable over long interactions, and learn from experience. Today's embodied systems largely follow an end-to-end observation-to-action path. Despite rapid progress, they remain far from this goal: their generalization depends heavily on the coverage of robot training data, while long task execution remains difficult to control and inspect. To realize this goal, we introduce the Embodied Task Agent (ETA), a new paradigm for extending digital agents into the physical world, and release OpenETA as its open-source implementation. ETA centers the robot around a Planner that chooses one Tool call at a time, an Interface that controls execution, and a World that returns the result and a fresh observation. This loop allows the agent to verify outcomes, adapt its plan, and turn successful and failed interactions into reusable experience. OpenETA provides replaceable Planners, composable Tools and Skills, auditable memory, replayable trajectories, and common interfaces for simulation and real robots. For Codex, OpenETA can operate as a lightweight plugin that exposes only observe, mark_point, and move_to.
Yitong Chen, Zezheng Huai, Sixian Li +7
1Shanghai Innovation Institute · 2Fudan University · 3Jilin University +2
Tool-augmented LLM agents commonly rely on step-wise atomic tool calls, where each invocation, observation, and value transfer is exposed in the main reasoning trace. This creates an execution-granularity mismatch: locally deterministic tool workflows are unfolded into repeated model-visible decisions, consuming context and forcing the model to manage low-level dataflow in the trace. We introduce HyperTool, a unified executable MCP-style tool interface that changes the model-visible unit of tool execution. A model invokes HyperTool with a code block that can call existing tools through their original schemas, manipulate returned values, and pass intermediate results locally, folding deterministic tool subroutines into a single outer call. To train models to use this interface, we synthesize HyperTool-format trajectories from cross-tool compositional tasks and verify them in real MCP environments. On MCP-Universe, HyperTool improves average accuracy from 15.69% to 35.29% on Qwen3-32B and from 9.93% to 33.33% on Qwen3-8B, and surpass GPT-OSS and Kimi-k2.5 on average accuracy, showing that our HyperTool can substantially improve multi-step tool use.
Yaxin Du, Yifan Zhou, Yujie Ge +7
Shanghai Jiao Tong University · IQuest Research · Beijing University of Aeronautics and Astronautics