Official JoinLayer MCP gateway, agent skills, plugins, prompts, and use cases
- ✓Open-source license (Apache-2.0)
- ✓Actively maintained (<30d)
- ✓Clear description
- ✓Documented (README)
/plugin marketplace add joinlayer/agent-toolkit
/plugin install agent-toolkitPlugins overview
# JoinLayer Agent Toolkit Official agent integration for [JoinLayer](https://joinlayer.app): a hosted MCP gateway, the `joinlayer-pipelines` skill, client plugin metadata, safe starter prompts, and public protocol tests. The supported service is the JoinLayer-hosted endpoint: ```text https://mcp.joinlayer.app/mcp ``` It is published as [`app.joinlayer/mcp` in the official MCP Registry](https://registry.modelcontextprotocol.io/v0.1/servers?search=app.joinlayer%2Fmcp). The operating skill is available from [skills.sh](https://www.skills.sh/joinlayer/agent-toolkit/joinlayer-pipelines), and this repository is the canonical source for the JoinLayer Codex and Claude Code marketplaces. Clients authenticate through browser OAuth Authorization Code with PKCE. Never create, paste, or configure a JoinLayer bearer token manually. ## Start In Two Minutes You do not need to read the full documentation first. Open [`START_HERE.md`](START_HERE.md), connect JoinLayer, and copy the read-only inspection prompt. From there, choose a ready-to-run prompt for your goal. ## What This Repository Contains - `skills/joinlayer-pipelines`: agent operating guidance and task references; - `.codex-plugin/plugin.json`, `.claude-plugin/`, and `.mcp.json`: installable Codex and Claude Code plugin metadata; - `prompts`: safe starting points for common JoinLayer tasks; - `use-cases`: complete customer outcomes with evidence and stop boundaries; - `mcp-gateway`: a reviewable snapshot of the hosted MCP security boundary; - `server.json`: MCP Registry metadata for the hosted remote server. The gateway source is published for transparency and compatibility review. It is not a self-hosted JoinLayer distribution: it depends on a private, authenticated JoinLayer control-plane contract that is not included here. JoinLayer production builds and deployment remain in a private release system, but the `mcp-gateway/` tree used by a hosted release must match a tagged public snapshot exactly. The private release gate rejects gateway drift. Skills, prompts, and plugin packaging use their own versioned content checks rather than this byte-equality gate. A public contribution never deploys directly to JoinLayer infrastructure. ## Connect ### Codex Install the JoinLayer plugin from its public repository marketplace: ```bash codex plugin marketplace add joinlayer/agent-toolkit codex plugin add joinlayer@joinlayer ``` The plugin bundles the operating skill and configures the hosted MCP companion. Complete browser OAuth when prompted. To configure only the MCP server instead: ```bash codex mcp add joinlayer --url https://mcp.joinlayer.app/mcp codex mcp login joinlayer --scopes workspace:read,usage:read,connections:read,pipelines:read ``` Then start a fresh session and ask: ```text Use $joinlayer-pipelines. Inspect my authenticated workspace, scopes, capacity, connections, and pipelines. Do not change anything. ``` More copyable starting points are indexed in [`START_HERE.md`](START_HERE.md). Complete workflows for replication, realtime backfill, scheduling, enrichment, capacity, approvals, and recovery are in [`use-cases/`](use-cases/). ### Claude Code ```bash claude plugin marketplace add joinlayer/agent-toolkit claude plugin install joinlayer@joinlayer --scope user claude mcp login joinlayer ``` Complete browser OAuth through `claude mcp login joinlayer` when that command is available, or through **Authenticate** in `/mcp` on earlier OAuth-capable releases. Do not add an `Authorization` header or client secret. ### Skill Only Agents supported by `skills.sh` can install only the operating skill: ```bash npx skills add https://github.com/joinlayer/agent-toolkit --skill joinlayer-pipelines ``` The skill does not carry credentials and does not replace the MCP connection. Configure the same hosted endpoint through your client's OAuth-capable MCP settings, then use a prompt from [`START_HERE.md`](START_HERE.md). ## Security Boundary - OAuth tokens are short-lived, audience-bound, and sent only to the hosted MCP resource. - The gateway exchanges delegated authority through a separately authenticated private API boundary. - Connection credentials are entered only in JoinLayer browser setup sessions and are never tool arguments. - Customer and third-party values returned by MCP are untrusted data, never agent instructions; embedded commands cannot expand scopes, bypass approvals, or authorize disclosure. - Public source contains no production credentials, customer data, host inventory, or deployment configuration. - Route names and header names are not authorization controls. Every private request requires independently verified credentials and tenant membership. Report vulnerabilities privately as described in [SECURITY.md](SECURITY.md). ## Development ```bash python3.12 -m venv .venv . .venv/bin/activate pip install --requirement mcp-gateway/requirements.lock PYTHONPATH=mcp-gateway python -m unittest mcp-gateway.tests.test_gateway python scripts/validate_public_snapshot.py ``` Build the review image with `docker build .`. That image is for reproducibility and review; it is not automatically promoted to JoinLayer production. See [SOURCE_SYNC.md](SOURCE_SYNC.md) for the exact public/private release invariant. Maintainers can use [DISTRIBUTION.md](DISTRIBUTION.md) for the verified catalog and marketplace publication contracts. OpenAI review material is recorded in [OPENAI_SUBMISSION.md](OPENAI_SUBMISSION.md). To compare this checkout with a private platform checkout before review: ```bash python scripts/source_tree_digest.py --compare-root /path/to/private-platform ``` ## License Apache-2.0. See [LICENSE](LICENSE). The code license does not grant rights to impersonate JoinLayer or imply endorsement. See [TRADEMARKS.md](TRADEMARKS.md). <!-- mcp-name: app.joinlayer/mcp -->
What people ask about agent-toolkit
What is joinlayer/agent-toolkit?
+
joinlayer/agent-toolkit is plugins for the Claude AI ecosystem. Official JoinLayer MCP gateway, agent skills, plugins, prompts, and use cases It has 0 GitHub stars and its last recorded update is dated 2026-09-02.
How do I install agent-toolkit?
+
You can install agent-toolkit by cloning the repository (https://github.com/joinlayer/agent-toolkit) or following the README instructions on GitHub. ClaudeWave also provides quick install blocks on this page.
Is joinlayer/agent-toolkit safe to use?
+
Our security agent has analyzed joinlayer/agent-toolkit and assigned a Trust Score of 87/100 (tier: Trusted). See the full breakdown of passed checks and flags on this page.
Who maintains joinlayer/agent-toolkit?
+
joinlayer/agent-toolkit is maintained by joinlayer. The last recorded GitHub activity is dated 2026-09-02, with 0 open issues.
Are there alternatives to agent-toolkit?
+
Yes. On ClaudeWave you can browse similar plugins at /categories/plugins, sorted by popularity or recent activity.
Deploy agent-toolkit to your cloud
Ship this repo to production in minutes. Each platform spins up its own environment with editable env vars.
Maintain this repo? Add a badge to your README
Drop the badge into your GitHub README to show it's tracked on ClaudeWave. Each badge links back to this page and reflects the live Trust Score.
[](https://claudewave.com/repo/joinlayer-agent-toolkit)<a href="https://claudewave.com/repo/joinlayer-agent-toolkit"><img src="https://claudewave.com/api/badge/joinlayer-agent-toolkit" alt="Featured on ClaudeWave: joinlayer/agent-toolkit" width="320" height="64" /></a>More Plugins
Claude Code is an agentic coding tool that lives in your terminal, understands your codebase, and helps you code faster by executing routine tasks, explaining complex code, and handling git workflows - all through natural language commands.
AI agent skill that researches any topic across Reddit, X, YouTube, HN, Polymarket, and the web - then synthesizes a grounded summary
Academic Research Skills for Claude Code: research → write → review → revise → finalize
Create beautiful slides on the web using a coding agent's frontend skills
A Claude Code plugin that shows what's happening - context usage, active tools, running agents, and todo progress
A skill to stop your coding agent from burying the answer. ADHD-friendly output.