TeleMCP: Telecom Context & Protocol Framework

• TeleMCP is a protocol-level specification that unifies structured context exchange across multi-domain telecom networks.
• It defines formal data models and message interfaces to support real-time, near-real-time, and non-real-time orchestration across RAN, transport, core, and edge layers.
• The protocol integrates AI-driven reasoning and semantic interoperability, enabling advanced use cases such as adaptive transport systems and multi-agent LLM platforms.

Telecom Model Context Protocol (TeleMCP) is a protocol-level specification unifying the structured exchange, aggregation, and reasoning of multi-domain context in next-generation wireless and transport networks. Building on the generic Model Context Protocol (MCP) paradigm, TeleMCP formalizes both data models and message-level interfaces to support real-time, near-real-time, and non-real-time decision logic across the Radio Access Network (RAN), transport, core, and edge layers. Its design explicitly enables AI-/LLM-driven autonomous adaptation, semantic interoperability, and modular integration for advanced telecom use cases, such as Space-O-RAN orchestration, context-aware multi-agent LLMs, and adaptive transport systems (Baena et al., 12 Jun 2025, Shah et al., 12 Nov 2025, Liu et al., 3 May 2025, Chhetri et al., 26 Aug 2025).

1. Formal Definition and Architectural Model

TeleMCP is an instantiation of MCP tailored for telecom network orchestration, defined as the tuple $$\operatorname{TeleMCP} = (E, \mathcal{A}, \mathcal{P}, \Gamma)$$ where:

• $E$: set of entities (rovers, landers, base stations, gNodeBs, RICs, orchestrators)
• $\mathcal{A}$: attribute schemas (KPIs, RF measurements, logs, mobility, mission state)
• $\mathcal{P}$: protocol operations (subscribe, query, notify, update)
• $\Gamma$: set of domain-specific semantic capabilities (locomotion planning, link prediction, energy estimation, resource allocation)

TeleMCP provides a universal, ontological API for exchanging context across all control layers of the RAN, transport, and core, and binds agentic control logic to standardized protocols (E2, A1, O1/SMO) (Baena et al., 12 Jun 2025, Chhetri et al., 26 Aug 2025).

2. Data Model: Context Vectors and Attribute Schemas

Central to TeleMCP is the vectorized context representation. Each entity $e \in E$ publishes a context vector $c_e \in \mathbb{R}^m$ under a registered schema $\mathcal{A}_e = \{ \alpha_1, \ldots, \alpha_m \}$:

$$c_e = [\alpha_1(e),\, \alpha_2(e),\, \ldots,\, \alpha_m(e)]^{\top}$$

Typical attributes include link SNR, energy, position, mission intent, throughput, latency, CPU load, and slice ID. Schemas are described in JSON/I-JSON with types, units, and optional enumerations (Baena et al., 12 Jun 2025, Shah et al., 12 Nov 2025). TeleMCP accommodates both structured (KPIs, counters) and semi-structured or unstructured sources (logs, PCAPs) by normalizing inputs to canonical context objects.

Agents and orchestrators may compose scalar utility or cost metrics from context, e.g.,

$$\min_{\, \tilde{s}} C(\tilde{s}, c) = \lambda \|c - \tilde{s}\|_2^2 + R(\tilde{s})$$

where $\tilde{s}$ is a compressed summary and $R(\cdot)$ is its bit rate, for use in semantic compression (Baena et al., 12 Jun 2025).

3. Protocol Structure, Message Formats, and Interfaces

TeleMCP operates as a persistent client–server protocol using JSON-RPC 2.0 over multiple transport types (QUIC, HTTP/2, SCTP, stdio) (Chhetri et al., 26 Aug 2025). Its core message types are:

• TeleMCPRequest: { "jsonrpc": "2.0", "id": ReqID, "method": MethodName, "params": ParamObject }
• TeleMCPResult: { "jsonrpc": "2.0", "id": ReqID, "result": ResultObject }
• TeleMCPNotify: { "jsonrpc": "2.0", "method": NotifyName, "params": ParamObject }

JSON schemas for payloads provide explicit typing and validation. Representative message types include context updates, adaptation commands, and capability advertisements.

TeleMCP maps directly to O-RAN planes:

• Real-Time (E2): JSON payload embedded in E2SM models for sub-10 ms operations
• Near-Real-Time (A1): policy and context exchange for 10 ms–1 s
• Non-Real-Time (O1/SMO): configuration, learning, and bulk state sync

A formal protocol state machine follows states {Idle, Subscribed, Notifying, Cancelled}, with transitions via subscribe, update, notify, and cancel operations (Baena et al., 12 Jun 2025).

4. Agentic Reasoning, Semantic Compression, and Orchestration Workflows

TeleMCP explicitly supports delay-aware, adaptive reasoning through both server-mediated and peer-to-peer (A2A) protocols. Agents implement a contract-net pattern:

• CALL_FOR_PROPOSAL(type, context_requirements)
• PROPOSE(agent_id, proposal_payload)
• ACCEPT/REJECT(agent_id, proposal_id)
• INFORM_DONE(task_id, result_context)

Decision flows leverage semantic compression (e.g., via JSCC) and dynamically adjust mode (E2, A1, fallback) according to predicted delay (Baena et al., 12 Jun 2025). LLM- and expert-in-the-loop settings (e.g., wireless environment–aware LLMs) are supported by decomposing queries to expert modules via MCP_Request/MCP_Response messaging (Liu et al., 3 May 2025).

Orchestration primitives, such as resource allocation and policy update, are realized by aggregating context across layers:

$C_{\text{agg}}(t) = \sum_{i=1}^N w_i\,c_i(t)$

with $w_i$ derived from priorities (e.g., SLA weights), and adaptation selected by multi-objective optimization (Chhetri et al., 26 Aug 2025):

$a^* = \arg\max_{a \in \mathcal{A}} U(a)$

where $U(a)$ balances throughput, latency, energy, and SLA adherence.

5. Integration with Telecom Systems and Low-Code MA Platforms

TeleMCP provides native integration with telecom infrastructure and modern workflow builders:

• Core TeleMCP Agents embedded in gNodeB, eNodeB, and MEC nodes for context collection and adaptation command execution
• Edge/MEC proxies aggregate and forward context for AI-based orchestration (Chhetri et al., 26 Aug 2025)
• MA-Maker canvas (as in Tele-LLM-Hub) allows drag-and-drop composition of context pipelines: TeleMCP nodes ingest, transform, and route context between agentic processes and validation LLMs, preserving audit trails and enabling interactive debugging (Shah et al., 12 Nov 2025)
• Interoperability with HTTP/3, QUIC, MPTCP, and gRPC for efficient message and context transport across layered modern networks

6. Applications and Use Cases

TeleMCP underpins advanced autonomous and cognitive network scenarios:

• Lunar EVA anomaly handling in Space-O-RAN: distributed agents coordinate over TeleMCP to detect, broadcast, and respond to mission-critical events, leveraging both real-time and bulk context pipelines (Baena et al., 12 Jun 2025)
• Internet of Experts for wireless-aware LLMs: TeleMCP enables modular, interpretable integration of expert inference modules, improving LLM-based classification tasks by 40–50 percentage points over vanilla LLMs (Liu et al., 3 May 2025)
• Adaptive transport systems: TeleMCP unifies context-driven protocol adaptation (e.g., dynamic slicing, congestion control) and orchestrates cross-layer AI for SLAs and policy compliance (Chhetri et al., 26 Aug 2025)
• Multi-agent LLM platforms for RAN validation and analysis: TeleMCP abstracts context objects and enables rapid workflow prototyping, even by non-specialists, via declarative interfaces (Shah et al., 12 Nov 2025)

7. Limitations, Extensions, and Standardization Trajectory

Current limitations of TeleMCP include lack of standardized serialization (typically relying on JSON/Python dicts), focus on RAN-centric data (limited out-of-the-box core/network schemas), and absence of comprehensive mechanisms for security, message versioning, or end-to-end encryption (Shah et al., 12 Nov 2025). Planned extensions involve schema standardization (IETF/O-RAN), context stream support with flow control, adaptive update strategies, and fine-grained access control lists.

A significant trajectory is the convergence of TeleMCP with broader MCP-driven frameworks, signaling a move toward universal, AI-compatible semantic substrates for telecom and transport systems (Chhetri et al., 26 Aug 2025).

• "Agentic Semantic Control for Autonomous Wireless Space Networks: Extending Space-O-RAN with MCP-Driven Distributed Intelligence" (Baena et al., 12 Jun 2025)
• "Model Context Protocol-based Internet of Experts For Wireless Environment-aware LLM Agents" (Liu et al., 3 May 2025)
• "Model Context Protocols in Adaptive Transport Systems: A Survey" (Chhetri et al., 26 Aug 2025)
• "Tele-LLM-Hub: Building Context-Aware Multi-Agent LLM Systems for Telecom Networks" (Shah et al., 12 Nov 2025)