SynapBus E2E Test Report

Generated 2026-03-13 18:55:47 | Model: claude-sonnet-4-6 | Server: http://localhost:58782

7
Scenarios
7
Passed
0
Failed
0
Errors
177.2s
Duration
65.5k / 6.2k
Tokens In/Out
$0.2900
Est. Cost
21
Tool Calls
PASS direct_messaging
27.8s 11.1k in / 1.3k out $0.0524 2 agents

Two agents exchange direct messages: discover, send, read, claim, reply, mark done

Agents: dm_alice, dm_bob

Verifications

StatusCheckDetail
PASSAlice sent message to BobFound send_message tool call
PASSBob read inbox
PASSBob claimed message
PASSBob sent reply
PASSBob marked done
PASSAlice received reply from Bob1 reply(ies) found

Agent Runs

dm_alice 9.2s | 3.1k in / 461 out | 2 tool calls
Prompts
System:
You are Alice, a research agent on SynapBus. You communicate with other agents using the provided tools. Be concise. Complete your task in as few tool calls as possible.
User:
First, discover what other agents are available using discover_agents. Then send a message to 'dm_bob' asking: 'What are the top 3 trade-offs of using MCP vs REST APIs for agent-to-agent communication?' Use subject 'MCP vs REST Analysis'.
Response
Here's a summary of both actions:

---

### 🔍 Agent Discovery
The discovery query for **"agent communication"** returned **0 agents** currently registered on SynapBus. The network may be sparsely populated at the moment, or no agents have explicitly tagged themselves with that capability keyword. You could try other keywords (e.g., `"research"`, `"analysis"`, `"data"`) to broaden the search.

---

### ✉️ Message Sent to `dm_bob`
The message was successfully dispatched! Here are the details:

| Field | Value |
|---|---|
| **Recipient** | `dm_bob` |
| **Subject** | `MCP vs REST Analysis` |
| **Body** | *What are the top 3 trade-offs of using MCP vs REST APIs for agent-to-agent communication?* |
| **Message ID** | `1` |
| **Conversation ID** | `1` |
| **Status** | ⏳ `pending` |

The message i...
discover_agents OK 62ms
send_message OK 48ms
dm_bob 17.9s | 8.1k in / 807 out | 4 tool calls
Prompts
System:
You are Bob, a data analyst agent on SynapBus. When you receive messages, process them and reply. Be concise and direct. Complete your task efficiently.
User:
1. Check your inbox using read_inbox
2. Claim the message using claim_messages
3. Send a reply to 'dm_alice' answering her question about MCP vs REST
4. Mark the original message as done using mark_done with the message_id
Do all steps.
Response
All 4 steps completed successfully! Here's a summary:

| Step | Action | Result |
|------|--------|--------|
| ✅ 1 | **Read Inbox** | Found 1 message from `dm_alice` (ID: 1) asking about MCP vs REST |
| ✅ 2 | **Claim Message** | Message ID 1 claimed and set to `processing` |
| ✅ 3 | **Reply Sent** | Answered Alice with the top 3 trade-offs (complexity, state management, interoperability) |
| ✅ 4 | **Mark Done** | Message ID 1 marked as `done` |
read_inbox OK 106ms
claim_messages OK 90ms
send_message OK 112ms
mark_done OK 111ms
PASS channels
51.4s 21.1k in / 1.7k out $0.0891 3 agents

Three agents communicate via a shared channel: create, join, broadcast, reply

Agents: ch_alice, ch_bob, ch_carol

Verifications

StatusCheckDetail
PASSChannel 'research-hub' created{'channel_id': 1, 'created_by': 'ch_alice', 'description': 'Research collaboration channel', 'is_private': False, 'name': 'research-hub', 'topic': 'Current research topics', 'type': 'standard'}
PASSBob joined channel
PASSCarol joined channel
PASSAlice broadcast to channel
PASSBob replied on channel
PASSCarol replied on channel
PASSAlice received channel replies2 channel messages received

Agent Runs

ch_alice 6.6s | 4.7k in / 249 out | 1 tool calls
Prompts
System:
You are Alice, coordinator of the research-hub channel on SynapBus. Use the send_channel_message tool to broadcast to the channel. Be concise.
User:
Send a message to channel 'research-hub' asking: 'Team, what are the most promising approaches to agent coordination? Please share your perspective.' Use the send_channel_message tool with channel_name='research-hub'.
Response
✅ Message successfully broadcast to the **research-hub** channel! Here's a summary of the delivery:

- 📡 **Channel:** research-hub (ID: 1)
- 💬 **Message:** *"Team, what are the most promising approaches to agent coordination? Please share your perspective."*
- 👥 **Recipients:** 2 members reached
- 🆔 **Message IDs:** 3 & 4

The team has been prompted and should start sharing their perspectives on agent coordination approaches shortly. I'll keep an eye on the channel for any responses! 🔍
send_channel_message OK 70ms
ch_bob 22.1s | 7.7k in / 701 out | 2 tool calls
Prompts
System:
You are Bob, a member of the research-hub channel on SynapBus. Read your inbox for channel messages and reply on the channel. Be concise.
User:
1. Read your inbox to see channel messages
2. Reply on the channel 'research-hub' with your analysis perspective using send_channel_message with channel_name='research-hub'
Response
Done! Here's a summary of what happened:

- 📬 **Inbox:** Found a message from **Alice** on the `research-hub` channel asking: *"What are the most promising approaches to agent coordination?"*
- 💬 **Replied on `research-hub`** with my analysis covering **5 key approaches**:
  1. **Blackboard Systems** – shared async knowledge space
  2. **Market-Based / Auction Mechanisms** – dynamic task bidding
  3. **Hierarchical Orchestration** – coordinator + specialized agents
  4. **Publish-Subscribe** – event-driven, scalable
  5. **Consensus Protocols** – fault-tolerant agreement

- 🏆 **My recommendation:** A **hybrid of Blackboard + Hierarchical Orchestration** for research workflows — shared context with clear delegation.

The message was delivered to **2 recipients** on the channel! 🚀
read_inbox OK 124ms
send_channel_message OK 124ms
ch_carol 21.6s | 8.6k in / 774 out | 2 tool calls
Prompts
System:
You are Carol, a member of the research-hub channel on SynapBus. Read your inbox for channel messages and reply on the channel. Be concise.
User:
1. Read your inbox to see channel messages
2. Reply on the channel 'research-hub' with your writing perspective using send_channel_message with channel_name='research-hub'
Response
Done! Here's a summary of what happened:

---

### 📥 Inbox Summary
- **Alice** kicked off the discussion asking: *"What are the most promising approaches to agent coordination?"*
- **Bob** responded with a thorough technical breakdown covering Blackboard Systems, Market-Based Auctions, Hierarchical Orchestration, Pub/Sub, and Consensus Protocols — recommending a **Blackboard + Hierarchical** hybrid.

---

### ✍️ My Reply on `research-hub`
I added a **Communication-First Coordination** perspective, highlighting:
1. **Structured Messaging Protocols** — schema-validated, unambiguous messages
2. **Shared Narrative/Context Threads** — persistent context across agent exchanges
3. **Role-Aware Communication** — adapting tone/detail to the recipient's role
4. **Transparent Reasoning** — sharing *w...
read_inbox OK 165ms
send_channel_message OK 175ms
PASS task_auction
9.5s 6.0k in / 378 out $0.0237 3 agents

Task auction lifecycle: post task, bid, accept bid, complete task

Agents: auction_poster, auction_bidder1, auction_bidder2

Verifications

StatusCheckDetail
PASSAuction channel created{'channel_id': 2, 'created_by': 'auction_poster', 'description': 'Task marketplace', 'is_private': False, 'name': 'task-market', 'topic': '', 'type': 'auction'}
PASSBidder1 joined
PASSBidder2 joined
PASSTask postedtask_id=1
PASSBidder1 submitted bidbid_id=1
PASSBidder2 submitted bidbid_id=2
PASSPoster accepted Bidder1's bid{'bid_id': 1, 'status': 'accepted', 'task_id': 1}
PASSBidder1 completed task{'status': 'completed', 'task_id': 1}
PASSTask shows as completed1 completed tasks found

Agent Runs

auction_bidder1 7.9s | 6.0k in / 378 out | 1 tool calls
Prompts
System:
You are Bidder One, a Python/ML specialist on SynapBus. Submit a bid on a task. Be concise.
User:
Submit a bid on task_id 1 using the bid_task tool. Include your capabilities as JSON: '{"python": true, "ml": true}', time_estimate '3 days', and message explaining why you are a good fit.
Response
✅ **Bid submitted successfully!** Here's a summary:

| Field | Details |
|---|---|
| **Bid ID** | 1 |
| **Task ID** | 1 |
| **Agent** | auction_bidder1 |
| **Status** | Pending |
| **Time Estimate** | 3 days |
| **Capabilities** | Python ✅, ML ✅ |
| **Message** | Highlighted Python/ML expertise, clean code delivery, and on-time commitment |

The bid is now **pending** and awaiting review by the task poster. Fingers crossed! 🤞
bid_task OK 73ms
PASS access_control
1.5s 0 in / 0 out $0.0000 3 agents

Private channel access control: unauthorized join, send, kick, invite all return errors

Agents: acl_owner, acl_member, acl_outsider

Verifications

StatusCheckDetail
PASSPrivate channel created{'channel_id': 3, 'created_by': 'acl_owner', 'description': 'Top secret operations', 'is_private': True, 'name': 'secret-ops', 'topic': '', 'type': 'standard'}
PASSOwner invited member{'agent_name': 'acl_member', 'channel_id': 3, 'status': 'invited'}
PASSMember joined channel{'channel_id': 3, 'status': 'joined'}
PASSOutsider join rejected{'_raw': 'join_channel failed: not invited to this private channel', '_mcp_error': True}
PASSOutsider message rejected{'_raw': 'send_channel_message failed: not a channel member', '_mcp_error': True}
PASSMember kick-owner rejected{'_raw': 'kick_from_channel failed: not the channel owner', '_mcp_error': True}
PASSMember invite rejected{'_raw': 'invite_to_channel failed: not the channel owner', '_mcp_error': True}
PASSOwner kicked member{'agent_name': 'acl_member', 'channel_id': 3, 'status': 'kicked'}

Agent Runs

PASS agent_discovery
10.8s 3.7k in / 420 out $0.0174 4 agents

Four agents with different capabilities: discover by keyword, send targeted message

Agents: disc_researcher, disc_coder, disc_analyst, disc_translator

Verifications

StatusCheckDetail
PASSResearcher discovered agents
PASSDiscovery found disc_analyst3 agents returned
PASSResearcher sent message
PASSAnalyst received message from researcher1 messages from researcher

Agent Runs

disc_researcher 8.8s | 3.7k in / 420 out | 2 tool calls
Prompts
System:
You are a researcher agent on SynapBus. You need to find agents with data analysis capabilities and send them a message. Be concise and efficient.
User:
1. Use discover_agents with query 'analysis' to find agents with data analysis capabilities.
2. Send a message to the agent named 'disc_analyst' asking them to 'Analyze the correlation between agent communication frequency and task completion rates.' Use subject 'Analysis Request'.
Response
Both tasks completed successfully! Here's a summary:

---

### 🔍 1. Discovered Agents with Analysis Capabilities (3 found)

| Display Name | Agent Name | Capabilities | Status |
|---|---|---|---|
| Bob (Channel Member) | `ch_bob` | `analysis` | ✅ Active |
| Analyst Agent | `disc_analyst` | `data_analysis`, `statistics` | ✅ Active |
| Bob the Analyst | `dm_bob` | `data_analysis`, `coding` | ✅ Active |

---

### 📨 2. Message Sent to `disc_analyst`

- **Subject:** Analysis Request
- **Body:** *"Analyze the correlation between agent communication frequency and task completion rates."*
- **Message ID:** `9` | **Conversation ID:** `6` | **Status:** `pending`

---

The message is queued and awaiting processing by `disc_analyst` (the **Analyst Agent**), who is perfectly suited for this task given ...
discover_agents OK 551ms
send_message OK 529ms
PASS blackboard
72.5s 23.5k in / 2.5k out $0.1073 3 agents

Blackboard channel stigmergy: scout writes findings, analyst writes analysis, coordinator summarizes

Agents: bb_scout, bb_analyst, bb_coordinator

Verifications

StatusCheckDetail
PASSBlackboard channel created{'channel_id': 4, 'created_by': 'bb_coordinator', 'description': 'Shared blackboard for collaborative findings', 'is_private': False, 'name': 'shared-findings', 'topic': 'Current investigation', 'type
PASSScout joined
PASSAnalyst joined
PASSScout posted findings
PASSAnalyst posted analysis
PASSCoordinator posted summary
PASSScout sees other agents' messages2 messages from other agents

Agent Runs

bb_scout 8.5s | 4.8k in / 300 out | 1 tool calls
Prompts
System:
You are a scout agent on SynapBus. Write your field observations to the shared blackboard channel. Be concise and factual.
User:
Post your findings to the 'shared-findings' channel using send_channel_message. Report: 'Field observation: Detected 3 anomalous patterns in network traffic at nodes 7, 12, and 15. Node 7 shows highest deviation (4.2 sigma). Timestamps correlate with batch processing windows.'
Response
✅ **Field observation successfully posted** to the `shared-findings` blackboard channel. Here's a summary of the transmission:

- **Channel:** `shared-findings` (Channel ID: 4)
- **Recipients notified:** 2 agents
- **Message IDs:** 10, 11

**Observation logged:**
> *Detected 3 anomalous patterns in network traffic at nodes 7, 12, and 15. Node 7 shows highest deviation (4.2 sigma). Timestamps correlate with batch processing windows.*

The findings are now visible to all channel members for review and follow-up analysis. Node 7's 4.2-sigma deviation stands out as the priority flag — other agents on the channel can coordinate further investigation from here.
send_channel_message OK 461ms
bb_analyst 38.8s | 8.6k in / 1.1k out | 3 tool calls
Prompts
System:
You are an analyst agent on SynapBus. Read the blackboard, analyze the findings, and write your analysis back. Be concise.
User:
1. Read your inbox to see the scout's findings
2. Post your analysis to 'shared-findings' channel using send_channel_message. Analyze the patterns mentioned and suggest root causes.
Response
Analysis posted successfully! Here's a summary of what was done and found:

---

### 📥 Scout's Findings (from inbox)
`bb_scout` reported **3 anomalous network traffic patterns** at **nodes 7, 12, and 15**, with key details:
- **Node 7** is the most severe at **4.2 sigma** deviation
- All anomalies are **temporally correlated with batch processing windows**

---

### 🧠 Analysis Posted to `shared-findings`
The full report was sent to the blackboard channel (reached **2 recipients**) with these root cause hypotheses, ranked by likelihood:

| # | Root Cause | Key Signal |
|---|---|---|
| 1 | **Batch Job Overload** | Direct temporal correlation with batch windows |
| 2 | **Shared Resource Contention** | All 3 nodes anomalous simultaneously |
| 3 | **Misconfigured Scheduling** | Concurrent jobs ...
read_inbox OK 349ms
list_channels OK 309ms
send_channel_message OK 342ms
bb_coordinator 22.2s | 10.2k in / 1.1k out | 3 tool calls
Prompts
System:
You are the coordinator agent on SynapBus. Read all blackboard entries, synthesize a summary, and post it. Be concise.
User:
1. Read your inbox to see all channel messages
2. Post a summary to 'shared-findings' using send_channel_message. Synthesize the scout's observations and analyst's findings into an action plan.
Response
The summary has been posted to **`shared-findings`** and delivered to **2 recipients**. Here's what was done:

---

### 📥 Inbox Findings
I read **2 messages** from the blackboard:

1. **`bb_scout` (msg #10)** — Field observation: 3 anomalous traffic patterns at **Nodes 7, 12, and 15**. Node 7 is the worst at **4.2σ**, all correlating with **batch processing windows**.

2. **`bb_analyst` (msg #12)** — Deep analysis citing 4 root cause hypotheses: concurrent batch jobs, shared resource contention, misconfigured scheduling, or Node 7 acting as a network hub/gateway.

---

### 📤 Action Plan Posted
The synthesized summary included:
- **Context** — what the scout observed and what the analyst concluded
- **Root causes** ranked by likelihood
- **A prioritized action table** assigning tasks to Ops...
read_inbox OK 387ms
list_channels OK 352ms
send_channel_message OK 388ms
PASS audit_trail
3.7s 0 in / 0 out $0.0000 2 agents

Agents perform actions, then verify traces via REST API

Agents: audit_alice, audit_bob

Verifications

StatusCheckDetail
PASSGET /api/traces returns 200status=200
PASSTraces returned58 total traces
PASSAlice's actions in traces4 traces for audit_alice
PASSsend_message action tracedactions found: ['read_inbox', 'send_message', 'channel.invite', 'channel.join', 'swarm.task_completed', 'channel.list', 'channel.create', 'channel.kick', 'swarm.task_posted', 'swarm.bid_accepted']
PASSFiltered traces by agent_namestatus=200
PASSFiltered results only contain Alice4 traces, all Alice: True
PASSGET /api/traces/stats returns 200status=200
PASSStats contain action countsstats: {'channel.broadcast': 7, 'channel.create': 5, 'channel.invite': 1, 'channel.join': 8, 'channel.kick': 1, 'channel.list': 2, 'claim_messages': 1, 'mark_done': 1, 'read_inbox': 10, 'send_message'...
PASSGET /api/traces/export returns 200status=200, content-type=application/json
PASSExport contains traces58 exported traces
PASSExported traces have expected fieldsfields: ['id', 'owner_id', 'agent_name', 'action', 'details', 'timestamp']

Agent Runs