Skip to main content

WxCC AI Agent Builder Toolkit (wxcc-toolkit)

Build AI agents and voice flows for Webex Contact Center with guided Claude Code assistance. This package delivers the entire Claude Code playbook — the wxcc-agent-builder agent, all 14 skills, all doc-lookup rules, and the full platform reference/playbook/template library — plus the wxcc-flow Flow Designer CLI, and materializes them into a working folder with one command.

Install

pip install wxcc-toolkit

This installs two console commands (aliases of the same CLI): wxcc-toolkit and wxcc-flow.

Create a project folder

wxcc-toolkit init my-wxcc-project

This materializes both the Claude Code and Codex CLI profiles into my-wxcc-project/ by default (pass --claude-only or --codex-only to install just one — see "Using with OpenAI Codex CLI" below):

my-wxcc-project/
├── CLAUDE.md                  # project instructions Claude Code loads at session start
├── AGENTS.md                  # project instructions Codex CLI loads
├── .mcp.json                  # Claude MCP server config (optional — see "Configure MCP servers" below)
├── .claude/
│   ├── settings.json          # pre-approved Skills, MCP tools, and wxcc-flow CLI
│   ├── agents/                # the wxcc-agent-builder agent
│   ├── skills/                # 14 build/design skills
│   └── rules/                 # platform-accuracy doc-lookup rules
├── .codex/
│   ├── config.toml            # approval policy, sandbox mode, MCP servers
│   ├── agents/                # per-agent Codex config
│   └── docs/                  # CLI command reference, doc-lookup rules
├── .agents/
│   └── skills/                # the same 14 skills, in Codex's skill format
└── docs/
    ├── reference/             # platform knowledge (Flow Designer, Connect, AI Agent Studio, …)
    ├── playbooks/             # step-by-step platform guides
    ├── templates/             # design-doc + import-JSON templates
    └── examples/              # working FlowIR / config examples

init writes a manifest per installed profile (.claude/.wxcc-manifest.json and/or .codex/.wxcc-manifest.json) recording every file it owns. It never overwrites files you create. If the target folder already contains files the playbook would overwrite, init stops and lists them — re-run with --force to overwrite, or pick an empty folder.

The wxcc-flow CLI is the preferred way to work with Flow Designer in this toolkit — 65 commands over the live 91-operation contract, pre-approved to run without permission prompts in the playbook's .claude/settings.json.

wxcc-flow configure

configure prompts for a Webex access token (generate a personal token at https://developer.webex.com — Getting Started → copy your token), then auto-resolves your org ID; the project ID resolves itself on first use. Developer-portal tokens expire after ~12 hours — re-run configure (or set WXCC_FLOW_TOKEN) when commands return a 401. The default endpoint is US-1 production; if your WxCC org is homed in a different region, pass your region's Flow Store base URL via wxcc-flow configure --base-url URL.

Configure MCP servers (optional)

Edit .mcp.json in your project folder and replace the placeholders. If you don't use one of these servers, remove its entry from .mcp.json.

Supabase (if you use Supabase as your database)

Flow Store (wxcc-flow-builder MCP — Cisco's official Flow Designer MCP server)

An alternative to the wxcc-flow CLI for Flow Designer work — this toolkit prefers the CLI, but Cisco's MCP server remains fully supported if you'd rather use it.

  • YOUR_FLOW_STORE_TOKEN — a Webex access token valid for your WxCC org. Generate a personal token at https://developer.webex.com (Getting Started → copy your token). Developer-portal tokens expire after ~12 hours; refresh and update .mcp.json when the server returns a 401.
  • The bundled URL (https://flow-store.produs1.ciscoccservice.com/flow-store/mcp) is the US-1 production endpoint. Use production endpoints only; if your WxCC org is homed in a different region, replace the host with your region's production Flow Store URL.

Start building

Open the folder in Claude Code and start a fresh session:

cd my-wxcc-project
claude

Then run /wxcc-agent-builder to build an agent, or /wxcc-debug to troubleshoot a failing action. The builder walks you through everything — from database setup through a working AI agent handling calls.

Using with OpenAI Codex CLI

The playbook ships both assistant profiles side by side. wxcc-flow init writes Claude Code files (CLAUDE.md, .claude/) and Codex files (AGENTS.md, .codex/, .agents/skills/) into the same folder — use --claude-only or --codex-only to restrict it:

pip install wxcc-toolkit
wxcc-flow init my-wxcc-project
cd my-wxcc-project
codex

Trust the folder when Codex prompts — the shipped .codex/config.toml (approval policy, sandbox mode, MCP servers) only takes effect in a trusted folder. AGENTS.md and the skills under .agents/skills/ load automatically; the CLI command reference lives in .codex/docs/cli-commands.md.

MCP servers (Flow Store token). Unlike Claude Code, the Codex Flow Store MCP server reads its token from an environment variable, not from a file — the generated .codex/config.toml sets bearer_token_env_var = "WXCC_FLOW_TOKEN". Export that variable (it is the same one wxcc-flow itself reads) so no token is ever written into a project file:

export WXCC_FLOW_TOKEN="your-webex-token"

wxcc-flow configure (used for the CLI) does NOT set this — export the env var yourself. If you also use the Supabase MCP server, replace the YOUR_SUPABASE_ACCESS_TOKEN / YOUR_SUPABASE_PROJECT_REF placeholders in .codex/config.toml.

The Codex profile is generated from the Claude profile at build time — the two never drift. Codex has no per-command allowlist, so expect an approval prompt when a command needs to run outside the workspace sandbox.

Update an existing project folder

After upgrading the package (pip install --upgrade wxcc-toolkit), refresh a project folder to the new playbook version:

wxcc-toolkit init my-wxcc-project --force

This automatically refreshes only the profiles already installed in that folder — a Claude-only folder stays Claude-only, a Codex-only folder stays Codex-only, and a dual-profile folder refreshes both. --force touches only manifest-owned files (retired skills are removed, updated docs are rewritten) and leaves any files you added untouched.

To add the other profile to an existing folder, pass its flag — wxcc-toolkit init my-wxcc-project --codex-only (or --claude-only).

Remove the playbook from a folder

wxcc-toolkit init my-wxcc-project --uninstall

Deletes every manifest-owned file and the manifest itself. Files you created are left in place.

For maintainers

src/wxcc_flow/_playbook/ is generated — never hand-edit it. After changing any shipped source (CLAUDE.md, .claude/agents|skills|rules, docs/reference|templates|playbooks|examples), regenerate the bundle:

python wxcc-dist/assemble.py

assemble.py enumerates tracked sources via git ls-files, substitutes the curated wxcc-dist/settings.bundled.json and sanitized wxcc-dist/mcp.bundled.json, and runs a link-audit gate that fails on residual dev-only paths. The release workflow re-runs it and refuses to publish if the committed bundle is stale.

Metadata

Release files for wxcc-toolkit 0.3.3

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for wxcc-toolkit 0.3.3
File Size Uploaded Provenance Actions
wxcc_toolkit-0.3.3.tar.gz 763.5 kB With provenance Details for wxcc_toolkit-0.3.3.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for wxcc-toolkit 0.3.3
File Interpreter ABI Platform Actions
wxcc_toolkit-0.3.3-py3-none-any.whl Python 3 none any Details for wxcc_toolkit-0.3.3-py3-none-any.whl

Total release size: 1.7 MB

Release files / wxcc_toolkit-0.3.3.tar.gz

Download URL wxcc_toolkit-0.3.3.tar.gz
Size 763.5 kB
Tags Source
SHA-256 checksum
How to use checksums
1f16601042f635a4ea6cef432a23b07efe06c3d715ef4bfcf444748b4419a294
BLAKE2b-256 checksum
How to use checksums
9bfd43dfda4e2c3fd6ebf5704f61eec0fd23683f34b694c516d3975badef11c4
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/6.1.0 CPython/3.13.12

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Jul 13, 2026.

Transparency log

Release files / wxcc_toolkit-0.3.3-py3-none-any.whl

Download URL wxcc_toolkit-0.3.3-py3-none-any.whl
Size 921.3 kB
Tags Python 3
SHA-256 checksum
How to use checksums
06ecbba48076a49907254afaceb9a1effbb1f9819ae50d2c3ecdfd954626f106
BLAKE2b-256 checksum
How to use checksums
d3b47ca53e4eb891096bdaa757ec5d9f208dc7d0e0defd83f53bfbef78d7af65
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/6.1.0 CPython/3.13.12

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Jul 13, 2026.

Transparency log

Security for wxcc-toolkit 0.3.3

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

Full provenance

Both files in this release have been uploaded with provenance metadata.

Select from 2 files

Attestation for wxcc_toolkit-0.3.3.tar.gz

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Jul 13, 2026.

Transparency log

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page