feat: publish TeleMem to the official MCP registry (v1.7.0)

- server.json manifest (io.github.teleai-uagi/telemem, pypi package,
  stdio transport, env var docs)
- mcp-name ownership marker in README (validated against the PyPI
  description by the registry)
- release.yml: mcp-registry job — installs mcp-publisher, syncs the
  manifest version to the tag, waits for PyPI visibility, authenticates
  via GitHub OIDC, publishes
- new `telemem` console script (alias of telemem-mcp) so `uvx telemem`
  runs the server zero-install, per the registry convention; mcp SDK
  moved into core deps to support it ([mcp] extra kept as alias)

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

Author: dell-zhang

.github/workflows/release.yml

@@ -56,3 +56,40 @@ jobs:
 
       - name: Publish to PyPI
         uses: pypa/gh-action-pypi-publish@release/v1
+
+  mcp-registry:
+    # Publish to the official MCP registry (registry.modelcontextprotocol.io).
+    # Auth: GitHub OIDC grants the io.github.teleai-uagi/* namespace; package
+    # ownership is proven by the "mcp-name:" marker in the PyPI description.
+    needs: publish
+    runs-on: ubuntu-latest
+    permissions:
+      id-token: write
+      contents: read
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Install mcp-publisher
+        run: |
+          curl -sL "https://github.com/modelcontextprotocol/registry/releases/latest/download/mcp-publisher_$(uname -s | tr '[:upper:]' '[:lower:]')_$(uname -m).tar.gz" | tar xz mcp-publisher
+
+      - name: Sync server.json version to tag
+        run: |
+          TAG_VERSION="${GITHUB_REF_NAME#v}"
+          jq --arg v "$TAG_VERSION" '.version = $v | .packages[0].version = $v' server.json > server.json.tmp
+          mv server.json.tmp server.json
+
+      - name: Wait for PyPI to serve the new version
+        run: |
+          TAG_VERSION="${GITHUB_REF_NAME#v}"
+          for i in $(seq 1 30); do
+            curl -sf "https://pypi.org/pypi/telemem/$TAG_VERSION/json" >/dev/null && exit 0
+            sleep 10
+          done
+          echo "PyPI version $TAG_VERSION not visible after 5 minutes" >&2
+          exit 1
+
+      - name: Publish to MCP registry
+        run: |
+          ./mcp-publisher login github-oidc
+          ./mcp-publisher publish

CHANGELOG.md

@@ -3,9 +3,14 @@
 All notable changes to TeleMem are documented here.
 The format follows [Keep a Changelog](https://keepachangelog.com/en/1.1.0/).
 
-## [Unreleased]
+## [1.7.0] - 2026-06-12
 
 ### Added
+- Published to the official [MCP registry](https://registry.modelcontextprotocol.io)
+  as `io.github.teleai-uagi/telemem`: `server.json` manifest, PyPI ownership
+  marker, and an OIDC-authenticated publish job in the release workflow.
+- New `telemem` console script (alias of `telemem-mcp`) so `uvx telemem`
+  runs the MCP server with zero install, per the registry convention.
 - Evaluation charter (`docs/evaluation.md`) aligned with
   [*The Benchmark Theatre*](https://essays.bloo-mind.ai/posts/2026-05-20-mem-eval/):
   baselines before architecture, constant base model, deterministic scoring,
@@ -17,6 +22,10 @@ The format follows [Keep a Changelog](https://keepachangelog.com/en/1.1.0/).
   The charter applies to new runs; existing published results are unchanged
   pending team review.
 
+### Changed
+- The MCP SDK (`mcp>=1.6.0`) is now a core dependency — TeleMem is
+  MCP-ready out of the box; `telemem[mcp]` remains as a no-op alias.
+
 ## [1.6.0] - 2026-06-12
 
 First release published to PyPI: `pip install telemem`.

README-ZH.md

@@ -565,11 +565,12 @@ print(f"Answer: ({answer})")
 TeleMem 内置 [Model Context Protocol](https://modelcontextprotocol.io)（MCP）服务器，任何兼容 MCP 的客户端——Claude Desktop、Claude Code、Cursor、自定义 Agent——都可以把 TeleMem 用作长期记忆。
 
 ```shell
-pip install "telemem[mcp]"
+pip install telemem
 
 telemem-mcp                                      # stdio（默认）
 telemem-mcp --transport sse --port 8421          # SSE over HTTP
 TELEMEM_CONFIG=config/config.yaml telemem-mcp    # 自定义 TeleMem 配置
+uvx telemem                                      # 零安装运行（stdio）
 ```
 
 服务器提供八个工具：`add_memory`、`search_memories`、`get_memories`、`get_memory`、`update_memory`、`delete_memory`、`delete_all_memories` 和 `memory_history`。未显式指定作用域的调用默认使用 `TELEMEM_DEFAULT_USER_ID`（`telemem-mcp`）；批量删除等破坏性操作必须显式指定作用域。

README.md

@@ -582,11 +582,12 @@ print(f"Answer: ({answer})")
 TeleMem ships a [Model Context Protocol](https://modelcontextprotocol.io) (MCP) server, so any MCP-compatible client — Claude Desktop, Claude Code, Cursor, custom agents — can use TeleMem as its long-term memory.
 
 ```shell
-pip install "telemem[mcp]"
+pip install telemem
 
 telemem-mcp                                      # stdio (default)
 telemem-mcp --transport sse --port 8421          # SSE over HTTP
 TELEMEM_CONFIG=config/config.yaml telemem-mcp    # custom TeleMem config
+uvx telemem                                      # zero-install run (stdio)
 ```
 
 The server exposes eight tools: `add_memory`, `search_memories`, `get_memories`, `get_memory`, `update_memory`, `delete_memory`, `delete_all_memories`, and `memory_history`. Calls without an explicit scope default to `TELEMEM_DEFAULT_USER_ID` (`telemem-mcp`); destructive bulk deletion always requires an explicit scope.
@@ -754,4 +755,6 @@ Made with ❤️ by [Bloo-Mind AI Ltd](https://www.bloo-mind.ai/) and the Ubiqui
     <a href="https://www.bloo-mind.ai/"><img src="assets/bloo-mind.png" alt="Bloo-Mind Logo" width="120px" /></a>
     &nbsp;&nbsp;&nbsp;
     <img src="assets/TeleAI.png" alt="TeleAI Logo" width="120px" />
-</div> 
+</div>
+
+<sub>mcp-name: io.github.teleai-uagi/telemem</sub>

pyproject.toml

@@ -20,9 +20,12 @@ dependencies = [
     "numpy",
     "pyyaml",
     "tqdm",
+    "mcp>=1.6.0",
 ]
 
 [project.optional-dependencies]
+# Kept as an alias for backward compatibility: the MCP SDK is now a core
+# dependency so `uvx telemem` works without extras (MCP registry convention).
 mcp = [
     "mcp>=1.6.0"
 ]
@@ -44,6 +47,9 @@ dev = [
 
 [project.scripts]
 telemem-mcp = "telemem.mcp.server:main"
+# `telemem` == the MCP server on stdio, so `uvx telemem` works out of the box
+# (the MCP registry convention is that the package identifier is runnable).
+telemem = "telemem.mcp.server:main"
 
 [project.urls]
 Homepage = "https://github.com/TeleAI-UAGI/telemem"

server.json

@@ -0,0 +1,42 @@
+{
+  "$schema": "https://static.modelcontextprotocol.io/schemas/2025-09-29/server.schema.json",
+  "name": "io.github.teleai-uagi/telemem",
+  "description": "Long-term and multimodal memory for AI agents - character-aware, mem0-compatible, fully-local option",
+  "repository": {
+    "url": "https://github.com/TeleAI-UAGI/telemem",
+    "source": "github"
+  },
+  "version": "1.7.0",
+  "packages": [
+    {
+      "registryType": "pypi",
+      "registryBaseUrl": "https://pypi.org",
+      "identifier": "telemem",
+      "version": "1.7.0",
+      "runtimeHint": "uvx",
+      "transport": {
+        "type": "stdio"
+      },
+      "environmentVariables": [
+        {
+          "name": "OPENAI_API_KEY",
+          "description": "API key for the default OpenAI LLM/embedder (not needed with a local TELEMEM_CONFIG)",
+          "isRequired": false,
+          "isSecret": true
+        },
+        {
+          "name": "TELEMEM_CONFIG",
+          "description": "Path to a TeleMem YAML/JSON config, e.g. the fully-local Ollama + FAISS setup (config/config.ollama.yaml)",
+          "isRequired": false,
+          "isSecret": false
+        },
+        {
+          "name": "TELEMEM_DEFAULT_USER_ID",
+          "description": "Default memory scope when tool calls omit user_id/agent_id/run_id (default: telemem-mcp)",
+          "isRequired": false,
+          "isSecret": false
+        }
+      ]
+    }
+  ]
+}

telemem/__init__.py

@@ -7,7 +7,7 @@
 from .mem0 import TeleMemory
 from .mem0 import TeleMemory as Memory
 
-__version__ = "1.6.0"
+__version__ = "1.7.0"
 
 # Everything mem0 exposes, with TeleMem's classes layered on top.
 __all__ = sorted(