Skip to content

Commit ae28be8

Browse files
shabarirepo2shabarirepo2
committed
Add Bonito AI infrastructure plugin
Bonito is a unified AI gateway for managing multi-provider infrastructure. This plugin gives Claude the domain knowledge to help teams: - Connect cloud AI providers (AWS Bedrock, Azure OpenAI, GCP Vertex, OpenAI, Anthropic, Groq) - Deploy infrastructure from declarative YAML configs - Create BonBon agents and Bonobot orchestrators - Configure intelligent routing (failover, cost-optimized, A/B testing) - Analyze and optimize AI spending across providers - Debug gateway, provider, and agent issues Includes 6 skills and an MCP server integration (bonito-mcp on PyPI/Docker Hub).
1 parent d2ba7f6 commit ae28be8

File tree

11 files changed

+672
-0
lines changed

11 files changed

+672
-0
lines changed

‎partner-built/bonito/.claude-plugin/plugin.json‎

Lines changed: 13 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,13 @@
1+
{
2+
"name": "bonito",
3+
"version": "1.0.0",
4+
"description": "Deploy and manage AI infrastructure across cloud providers. Connect AWS Bedrock, Azure OpenAI, GCP Vertex AI, OpenAI, Anthropic, and Groq through a unified gateway. Create agents, configure intelligent routing with failover and cost optimization, analyze spending, and debug issues.",
5+
"author": {
6+
"name": "Bonito AI",
7+
"url": "https://getbonito.com"
8+
},
9+
"repository": "https://github.com/ShabariRepo/bonito",
10+
"homepage": "https://getbonito.com",
11+
"license": "MIT",
12+
"keywords": ["ai-infrastructure", "multi-provider", "routing", "agents", "gateway", "aws-bedrock", "azure-openai", "gcp-vertex", "cost-optimization", "failover"]
13+
}

‎partner-built/bonito/.mcp.json‎

Lines changed: 10 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,10 @@
1+
{
2+
"mcpServers": {
3+
"bonito": {
4+
"command": "bonito-mcp",
5+
"env": {
6+
"BONITO_API_KEY": ""
7+
}
8+
}
9+
}
10+
}

‎partner-built/bonito/CONNECTORS.md‎

Lines changed: 54 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,54 @@
1+
# Connectors
2+
3+
## How tool references work
4+
5+
Plugin files use `~~category` as a placeholder for whatever tool the user connects in that category. For example, `~~bonito` refers to the Bonito MCP server, which provides access to the Bonito AI Gateway API.
6+
7+
Plugins are **tool-agnostic** — they describe workflows in terms of categories (gateway, notifications, version control, etc.) rather than specific products. The `.mcp.json` pre-configures the Bonito MCP server, but optional connectors in other categories enhance the experience.
8+
9+
## Connectors for this plugin
10+
11+
| Category | Placeholder | Included servers | Other options |
12+
|----------|-------------|-----------------|---------------|
13+
| AI Gateway | `~~bonito` | Bonito MCP | — |
14+
| Version Control | `~~vcs` | — | GitHub, GitLab |
15+
| Notifications | `~~notifications` | — | Slack, Microsoft Teams, Discord |
16+
| Monitoring | `~~monitoring` | — | Datadog, Grafana, PagerDuty |
17+
18+
## Required connector
19+
20+
### Bonito MCP (`~~bonito`)
21+
22+
The core connector. Provides access to the Bonito AI Gateway API for managing providers, agents, routing, and infrastructure.
23+
24+
**Install:** `pip install bonito-mcp`
25+
26+
**What it provides:**
27+
- Provider management (create, list, verify, delete)
28+
- Agent configuration (BonBon agents, Bonobot orchestrators)
29+
- Routing policy management (failover, cost-optimized, A/B testing)
30+
- Cost and usage analytics
31+
- Gateway health and diagnostics
32+
- Knowledge base management
33+
34+
## Optional connectors
35+
36+
### GitHub (`~~vcs`)
37+
38+
Connect GitHub to enable infrastructure-as-code workflows, review bonito.yaml changes in PRs, and track deployment history.
39+
40+
**What it adds:**
41+
- Read bonito.yaml configs from repos
42+
- Review infrastructure changes in pull requests
43+
- Track deployment commits and history
44+
- Manage infrastructure configs alongside application code
45+
46+
### Slack (`~~notifications`)
47+
48+
Connect Slack to receive deployment notifications, cost alerts, and provider status updates in your team channels.
49+
50+
**What it adds:**
51+
- Deployment success/failure notifications
52+
- Cost threshold alerts
53+
- Provider health status updates
54+
- Agent error notifications

‎partner-built/bonito/LICENSE‎

Lines changed: 21 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,21 @@
1+
MIT License
2+
3+
Copyright (c) 2026 Bonito AI
4+
5+
Permission is hereby granted, free of charge, to any person obtaining a copy
6+
of this software and associated documentation files (the "Software"), to deal
7+
in the Software without restriction, including without limitation the rights
8+
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
9+
copies of the Software, and to permit persons to whom the Software is
10+
furnished to do so, subject to the following conditions:
11+
12+
The above copyright notice and this permission notice shall be included in all
13+
copies or substantial portions of the Software.
14+
15+
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16+
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17+
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
18+
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19+
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
20+
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
21+
SOFTWARE.

‎partner-built/bonito/README.md‎

Lines changed: 108 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,108 @@
1+
# Bonito Plugin
2+
3+
An AI infrastructure management plugin primarily designed for [Cowork](https://claude.com/product/cowork), Anthropic's agentic desktop application — though it also works in Claude Code. Helps you deploy and manage AI infrastructure across cloud providers, create agents, configure intelligent routing, analyze costs, and debug issues. Works with any team — standalone with the Bonito API, supercharged when you connect GitHub, Slack, and other tools.
4+
5+
## Installation
6+
7+
```bash
8+
claude plugin install bonito
9+
```
10+
11+
## Skills
12+
13+
Domain knowledge Claude uses automatically when relevant:
14+
15+
| Skill | Description |
16+
|---|---|
17+
| `deploy-stack` | Deploy AI infrastructure from a bonito.yaml config — create providers, agents, knowledge bases, and routing in one shot |
18+
| `manage-providers` | Connect, verify, and manage cloud AI providers — AWS Bedrock, Azure OpenAI, GCP Vertex AI, OpenAI, Anthropic, Groq |
19+
| `create-agent` | Create and configure BonBon agents or Bonobot orchestrators with system prompts, models, MCP tools, and RAG |
20+
| `cost-analysis` | Analyze AI spending across providers, identify expensive models, recommend cheaper alternatives, optimize routing |
21+
| `gateway-routing` | Configure routing policies — cost-optimized, failover, A/B testing, model aliases, and cross-region inference |
22+
| `debug-issues` | Troubleshoot gateway errors, provider failures, and agent issues — check logs, verify connections, test endpoints |
23+
24+
## Example Workflows
25+
26+
### Deploying Your AI Stack
27+
28+
Just describe what you want:
29+
```
30+
Deploy my AI infrastructure from bonito.yaml
31+
```
32+
33+
The `deploy-stack` skill reads your config file, creates providers, agents, knowledge bases, and routing policies — all in one shot. It validates each step and reports what was created.
34+
35+
### Connecting a New Provider
36+
37+
```
38+
Connect my AWS Bedrock account as a provider
39+
```
40+
41+
The `manage-providers` skill walks you through credential setup, creates the provider, verifies the connection, and lists available models. Works for any supported cloud provider.
42+
43+
### Building an Agent
44+
45+
```
46+
Create a customer support agent using Claude on AWS Bedrock with our FAQ knowledge base
47+
```
48+
49+
The `create-agent` skill configures a BonBon agent with your chosen model, system prompt, knowledge base, and MCP tools. It deploys the agent and gives you the endpoint.
50+
51+
### Optimizing Costs
52+
53+
```
54+
What am I spending on AI across all providers?
55+
```
56+
57+
The `cost-analysis` skill pulls usage data across all connected providers, breaks down costs by model and agent, identifies expensive patterns, and recommends cheaper alternatives or routing changes.
58+
59+
## Standalone + Supercharged
60+
61+
Every skill works without any additional integrations:
62+
63+
| What You Can Do | Standalone | Supercharged With |
64+
|-----------------|------------|-------------------|
65+
| Deploy infrastructure | bonito.yaml + Bonito API | GitHub (track configs in repos) |
66+
| Manage providers | Bonito API | Monitoring (health dashboards) |
67+
| Create agents | Bonito API | GitHub (version agent configs) |
68+
| Analyze costs | Bonito API | Slack (cost threshold alerts) |
69+
| Configure routing | Bonito API | Monitoring (traffic dashboards) |
70+
| Debug issues | Bonito API logs | Slack (error notifications), Monitoring |
71+
72+
## MCP Integrations
73+
74+
> If you see unfamiliar placeholders or need to check which tools are connected, see [CONNECTORS.md](CONNECTORS.md).
75+
76+
Connect your tools for a richer experience:
77+
78+
| Category | Examples | What It Enables |
79+
|---|---|---|
80+
| **AI Gateway** | Bonito MCP | Provider management, agent creation, routing, cost analytics |
81+
| **Version Control** | GitHub, GitLab | Infrastructure-as-code, config versioning, PR reviews |
82+
| **Notifications** | Slack, Teams, Discord | Deployment alerts, cost warnings, provider status updates |
83+
| **Monitoring** | Datadog, Grafana | Traffic dashboards, latency tracking, health monitoring |
84+
85+
See [CONNECTORS.md](CONNECTORS.md) for the full list of supported integrations.
86+
87+
## Settings
88+
89+
Create a `settings.local.json` file to personalize:
90+
91+
- **Cowork**: Save it in any folder you've shared with Cowork (via the folder picker). The plugin finds it automatically.
92+
- **Claude Code**: Save it at `bonito/.claude/settings.local.json`.
93+
94+
```json
95+
{
96+
"organization": "Your Company",
97+
"bonito_api_url": "https://api.getbonito.com",
98+
"default_model": "anthropic/claude-sonnet-4-20250514",
99+
"preferences": {
100+
"default_provider": "aws-bedrock",
101+
"cost_alert_threshold": 100,
102+
"routing_strategy": "cost-optimized",
103+
"region": "us-east-1"
104+
}
105+
}
106+
```
107+
108+
The plugin will ask you for this information interactively if it's not configured.

‎partner-built/bonito/skills/cost-analysis/SKILL.md‎

Lines changed: 75 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,75 @@
1+
---
2+
name: cost-analysis
3+
description: "Analyze AI spending across providers, identify expensive models, recommend cheaper alternatives, and optimize routing for cost savings. Triggers on 'what am I spending', 'cost breakdown', 'optimize my AI costs', 'which models cost the most', or 'reduce my AI costs'."
4+
---
5+
6+
# Cost Analysis
7+
8+
Understand and optimize AI spending across all connected providers. Pull usage data, break down costs, identify expensive patterns, and recommend concrete savings.
9+
10+
## Step 1: Determine Scope
11+
12+
- "What am I spending?" -> full overview, current month
13+
- "Costs by provider" -> provider breakdown
14+
- "Most expensive models" -> model ranking
15+
- "Compare months" -> trend analysis
16+
- "Optimize costs" -> recommendations focus
17+
- "Last 7 days" -> custom date range
18+
19+
## Step 2: Pull Usage Data
20+
21+
Query ~~bonito for:
22+
1. Total requests, tokens, and costs for the period
23+
2. Breakdown by provider
24+
3. Breakdown by model
25+
4. Breakdown by agent
26+
5. Daily time series for trends
27+
28+
## Step 3: Analyze Patterns
29+
30+
- Which models account for the most spending?
31+
- Are agents using expensive models for simple tasks?
32+
- Is usage concentrated or spread across providers?
33+
- Are there traffic spikes or retry storms?
34+
- Is token usage efficient (prompt length vs response length)?
35+
36+
## Step 4: Generate Recommendations
37+
38+
**Model substitution:**
39+
- Opus for simple tasks -> suggest Haiku or Sonnet
40+
- GPT-4o for classification -> suggest GPT-4o-mini
41+
- High-volume agent -> suggest Groq for speed + cost
42+
43+
**Routing optimization:**
44+
- One provider consistently more expensive -> route to cheaper alternative
45+
- Bursty traffic -> suggest response caching
46+
- Quality varies -> A/B test cheaper models
47+
48+
## Step 5: Present Report
49+
50+
```
51+
## AI Cost Analysis
52+
53+
Period: [Start] to [End]
54+
Total Spend: $[Amount] | Requests: [Count] | Tokens: [Count]
55+
56+
### By Provider
57+
| Provider | Requests | Cost | % of Total |
58+
|----------|----------|------|------------|
59+
| AWS Bedrock | [Count] | $[Amount] | [X]% |
60+
| OpenAI | [Count] | $[Amount] | [X]% |
61+
62+
### By Model (Top 5)
63+
| Model | Provider | Requests | Cost | $/1K Req |
64+
|-------|----------|----------|------|----------|
65+
| claude-sonnet | Bedrock | [Count] | $[Amount] | $[Amount] |
66+
67+
### Recommendations
68+
1. [Action]: Switch [agent] from [expensive model] to [cheaper model]
69+
Estimated savings: $[Amount]/month ([X]%)
70+
71+
2. [Action]: [Description]
72+
Estimated savings: $[Amount]/month ([X]%)
73+
74+
Potential monthly savings: $[Amount] ([X]% reduction)
75+
```

‎partner-built/bonito/skills/create-agent/SKILL.md‎

Lines changed: 82 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,82 @@
1+
---
2+
name: create-agent
3+
description: "Create and configure BonBon agents or Bonobot orchestrators with system prompts, models, MCP tools, and RAG knowledge bases. Triggers on 'create an agent', 'deploy a chatbot', 'set up a support bot', 'build an orchestrator', 'make a BonBon', or 'create a Bonobot'."
4+
---
5+
6+
# Create Agent
7+
8+
Create AI agents on the Bonito gateway. Handles both BonBon agents (single-model, task-focused) and Bonobot orchestrators (multi-agent, tool-using).
9+
10+
## Step 1: Determine Agent Type
11+
12+
- Simple task, single model, "chatbot", "support bot" -> **BonBon**
13+
- Multi-step, coordination, "orchestrator", "multi-agent" -> **Bonobot**
14+
15+
If unclear, ask: "Do you need a single-purpose agent or a multi-agent orchestrator?"
16+
17+
## Step 2: Select Model and Provider
18+
19+
1. List available providers via ~~bonito
20+
2. Recommend model based on use case:
21+
- Customer support -> Claude Sonnet (good balance)
22+
- Complex reasoning -> Claude Opus or GPT-4o
23+
- High volume, simple tasks -> Claude Haiku or Groq Llama
24+
- Code generation -> Claude Sonnet or GPT-4o
25+
3. Confirm with user
26+
27+
## Step 3: Configure System Prompt
28+
29+
If user provides one, use it directly. If they describe the role:
30+
1. Generate a draft (role, tone, boundaries, output format)
31+
2. Present for approval
32+
3. Iterate until satisfied
33+
34+
## Step 4: Attach Knowledge Base (Optional)
35+
36+
1. Check for existing KBs via ~~bonito
37+
2. If new: create KB, upload documents, wait for indexing
38+
3. Configure retrieval settings (top-K, similarity threshold)
39+
40+
## Step 5: Configure MCP Tools (Optional)
41+
42+
Select and attach relevant MCP tool servers for the agent's purpose.
43+
44+
## Step 6: Create, Deploy, and Test
45+
46+
1. Assemble config and call ~~bonito to create
47+
2. Send a basic greeting -> verify response
48+
3. Send a domain query -> verify relevance
49+
4. If tools/KB attached -> test those too
50+
51+
## Output Format
52+
53+
```
54+
## Agent Created: [Name]
55+
56+
| Field | Value |
57+
|-------|-------|
58+
| Type | [BonBon / Bonobot] |
59+
| Model | [Model] via [Provider] |
60+
| Agent ID | [ID] |
61+
| Endpoint | /v1/agents/[id]/chat |
62+
| Knowledge Base | [Name or None] |
63+
| MCP Tools | [Count or None] |
64+
| Status | Deployed |
65+
66+
### Test Results
67+
| Test | Status |
68+
|------|--------|
69+
| Basic greeting | OK |
70+
| Domain query | OK |
71+
72+
### Quick Start
73+
curl -X POST https://api.getbonito.com/v1/agents/[id]/chat \
74+
-H "Authorization: Bearer $BONITO_API_KEY" \
75+
-d '{"message": "Hello!"}'
76+
```
77+
78+
## Agent Types
79+
80+
**BonBon:** Single-model agent for focused tasks (support bots, Q&A, code review, content generation). One model, optional system prompt, optional KB, optional MCP tools.
81+
82+
**Bonobot:** Multi-agent orchestrator that delegates to specialized sub-agents. For complex workflows spanning multiple domains or requiring different models for different tasks.

0 commit comments

Comments
 (0)