Multi-Agent Bridge
Bridge can act as an orchestrator — not just using its own built-in reasoning, but also delegating to your own agents. When you connect one of your project agents to Bridge, Bridge can call that agent as a specialized delegate, using your own LLM key and your own agent's specific expertise.
Why Use Agents in Bridge?
Bridge's built-in model is great for general orchestration and project management. But sometimes you need:
- Domain expertise — Your custom-configured analyst, code reviewer, or content writer agent has specific skills and RAG sources that Bridge doesn't have
- Specific models — Your task might need a particular model's larger context window or stronger reasoning
- Specialized tools — Your agents have web_search, web_scraper, or MCP tools configured
- Cost control — Use the free built-in model for orchestration, only spinning up your own (paid) models for the tasks that need them
Connecting an Agent to Bridge
Method 1: During Conversation
Type a message asking Bridge to involve an agent:
You: "I need you to analyze this security log. Use the security-analyst agent for the analysis."
Bridge: I can involve your security-analyst agent in this conversation.
They'll handle the security log analysis using their specialized configuration.
Would you like to add security-analyst to this conversation?
[✓ Add security-analyst] [✗ No thanks]
Click Add security-analyst — the agent joins the conversation.
Method 2: Agent Picker
In the Bridge header or conversation panel, click + Agent (person icon). A dropdown shows all your project agents. Select one to add it to the conversation. You can add multiple agents.
Agent Join Flow
When an agent joins the conversation:
- Bridge shows an agent badge in the conversation header (e.g., 🔬 security-analyst)
- A join message appears: "security-analyst has joined the conversation"
- Bridge can now delegate tasks to this agent
Bridge header: [Bridge 🤖] + [🔬 security-analyst]
Bridge: security-analyst is now available in this conversation.
They'll handle security analysis tasks with their specialized
configuration (model: claude-sonnet-4-6, RAG: company-security-docs).
Go ahead — what would you like to analyze?
How Delegation Works
Once an agent is in the conversation, Bridge decides when to delegate based on your request:
You: "Here are the logs from our firewall: [paste logs]
Analyze them for threats."
Bridge: [Delegates to security-analyst]
🔬 security-analyst:
Analysis of the provided firewall logs:
**Critical (1):**
- Multiple authentication failures from 192.168.1.105 — possible brute-force attempt.
Source IP made 47 failed login attempts in 3 minutes.
**High (2):**
- Unusual outbound traffic to 185.220.101.xx (known Tor exit node)
- Port scan detected from 10.0.1.55 — 500 unique ports in 2 minutes
**Recommended Actions:**
1. Block 192.168.1.105 immediately
2. Investigate 10.0.1.55 for compromise
...
The response is clearly attributed to security-analyst. Bridge summarizes or acts on the response as needed.
Token Costs for Agents in Bridge
When an agent is invoked in Bridge, your LLM key is used (the key configured on that agent). You pay your provider's rate for those calls — Bridge's built-in model is still free, but your agent's calls come from your account.
The cost shows in your provider's dashboard, associated with your API key.
Multiple Agents
You can add multiple agents to a single Bridge conversation. This lets you orchestrate complex multi-agent discussions:
Bridge header: [Bridge 🤖] + [🔬 security-analyst] + [💻 code-reviewer]
You: "The security analyst found a vulnerability in our auth code.
Have the code reviewer look at this code snippet and suggest fixes."
Bridge: [Delegates analysis to security-analyst, then code review to code-reviewer]
🔬 security-analyst: The JWT validation doesn't check the 'exp' claim...
💻 code-reviewer: Based on the security finding, here are the recommended fixes:
1. Add expiry validation: [code snippet]
2. Use constant-time comparison: [code snippet]
Removing Agents
To remove an agent from the conversation:
- Click the agent badge in the conversation header.
- Click Remove from conversation.
The agent's context from the conversation is removed. Subsequent messages won't involve that agent unless re-added.
Agent Availability
An agent must be configured in your project to be used in Bridge. If an agent uses an API key that belongs to another user, Bridge will show a warning (the key can't be accessed across users) and suggest you configure the agent with your own key.
Example: Full Multi-Agent Bridge Session
You: I need to process this week's security alerts and then fix the highest-severity code issue.
Bridge: I'll coordinate this. Let me add your security-analyst and code-reviewer agents.
[Adds both agents]
You: Here are this week's alerts: [paste 50 log lines]
Bridge: [Delegates to security-analyst]
🔬 security-analyst: [Analyzes logs, outputs JSON with severity classification]
Bridge: Security analyst found 3 critical issues and 7 high issues.
The most critical is a SQL injection vulnerability in the payment module.
Shall I have the code-reviewer look at the vulnerable code?
You: Yes, here's the relevant code: [paste code]
Bridge: [Delegates to code-reviewer]
💻 code-reviewer: I've reviewed the payment module code. The vulnerability is
on line 42 — the raw SQL query is built with string concatenation...
Here's the fix: [code with parameterized queries]
You: Perfect. Create a GitHub issue with the security analyst's findings and
the code reviewer's recommended fix.
Bridge: [Uses the GitHub MCP server to create the issue]
✅ Created GitHub issue #247: "Critical: SQL injection in payment module"
(link to the issue in your repo)
This entire flow — security analysis, code review, and GitHub issue creation — happens in one Bridge conversation, orchestrated automatically.