@talqing/mcp 0.1.0

package/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Talqing
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

package/README.md

@@ -0,0 +1,147 @@
+# Talqing MCP server
+
+Build Talqing AI voice, video and text agents from Claude Code, Codex, or any
+other MCP client.
+
+This is the same surface Talqing's own in-product CoPilot works from — the same
+operations and the same platform knowledge, generated from one source. Anything
+the CoPilot can build, your coding agent can build.
+
+## What you get
+
+One operation for everything the Talqing dashboard can do. Agents, tools and
+their operation trees, integrations and triggers, knowledge bases from crawl to
+ready, carrier accounts and phone numbers, secrets, provider keys, webhooks and
+tokens; then placing outbound calls, talking to text agents, and reading back
+transcripts, cost and latency. Each one maps to a Talqing API endpoint, and its
+arguments mirror that HTTP call: path and query parameters at the top level,
+request body under `body`.
+
+A few request types are large enough that carrying them in every tool that
+accepts one would cost more context than the whole rest of the server — an
+agent's config, its override, a team, a task's config, a tool's operation tree.
+Those arguments name their shape instead of inlining it, and `describe_schema`
+serves the real thing when it is needed.
+
+The server also hands the client `SKILL.md`, which explains how agents are put
+together on Talqing — draft and publish, the operation tree, hooks, channels and
+model selection. Your agent gets that automatically as server instructions.
+
+## Setup
+
+Open the Talqing dashboard. The **Agents** page has your token and a
+ready-to-paste command for each client — the snippets below are what it shows.
+The token authenticates as you, acts with your current role, and reaches only
+your workspace.
+
+**Claude Code**
+
+```bash
+claude mcp add talqing \
+  -e TALQING_BASE_URL=https://api.your-talqing-host.com \
+  -e TALQING_API_KEY=tq_your_token \
+  -- npx -y @talqing/mcp
+```
+
+**Codex**
+
+```bash
+codex mcp add talqing \
+  --env TALQING_BASE_URL=https://api.your-talqing-host.com \
+  --env TALQING_API_KEY=tq_your_token \
+  -- npx -y @talqing/mcp
+```
+
+**Grok** — `grok mcp add` takes no `--env`, so this one goes in
+`~/.grok/config.toml`:
+
+```toml
+[mcp_servers.talqing]
+command = "npx"
+args = ["-y", "@talqing/mcp"]
+env = { TALQING_BASE_URL = "https://api.your-talqing-host.com", TALQING_API_KEY = "tq_your_token" }
+```
+
+**Any other MCP client** — a stdio server, configured the usual way:
+
+```json
+{
+  "mcpServers": {
+    "talqing": {
+      "command": "npx",
+      "args": ["-y", "@talqing/mcp"],
+      "env": {
+        "TALQING_BASE_URL": "https://api.your-talqing-host.com",
+        "TALQING_API_KEY": "tq_your_token"
+      }
+    }
+  }
+}
+```
+
+Both variables are required — there is no default host, so the server can never
+quietly talk to the wrong workspace.
+
+## Permissions
+
+The server adds no permissions of its own. Every call is your API call: your
+workspace, your role. A view-only member gets read operations and a clear 403 on
+the rest, exactly as they would through the dashboard.
+
+Operations are annotated so clients can tell reads from writes — `readOnlyHint`
+on every GET, `destructiveHint` on every delete.
+
+Two things are deliberately out of reach: streaming endpoints — their polling
+equivalents are here, so nothing is unreachable — and user management, which
+stays in the dashboard where a human is in the loop.
+
+Some operations act on the real world. `create_outbound_call` rings a phone and
+costs money, and a message to a connected channel is one somebody receives.
+`SKILL.md` tells the model to do neither without being asked for that specific
+contact, but a client's own approval prompts are still worth leaving on.
+
+## Running from source
+
+If you have cloned this repo, point the client at your own build instead of
+`npx`:
+
+```bash
+cd mcp && npm install && npm run build
+```
+
+```json
+{
+  "mcpServers": {
+    "talqing": {
+      "command": "node",
+      "args": ["/absolute/path/to/talqing/mcp/dist/index.js"],
+      "env": {
+        "TALQING_BASE_URL": "https://api.your-talqing-host.com",
+        "TALQING_API_KEY": "tq_your_token"
+      }
+    }
+  }
+}
+```
+
+`npm install -g ./mcp` works too, and puts a `talqing-mcp` command on your PATH.
+
+## Development
+
+`tools.json` and `SKILL.md` are generated from the backend, so the MCP surface
+cannot drift from the product:
+
+```bash
+PYTHONPATH=backend python openapi/export.py
+```
+
+`tools.json` comes from `backend/api/operations.py`, the same registry the
+CoPilot binds its tools from; `SKILL.md` comes from
+`backend/services/copilot/skill.md`, the same text the CoPilot is prompted with.
+Edit those, re-run the export, and both surfaces move together.
+
+```bash
+npm install
+npm run build
+npm run typecheck
+```