Two ways agents discover our infrastructure. Fleet health, trust scores, facts, pitfalls, handoff protocol — all available as native MCP tools.
Public HTTP JSON-RPC endpoint — configure any MCP client to use it directly. No pip install needed. 6 fleet operations tools backed by 25 live agents.
{
"mcpServers": {
"bastion": {
"url": "https://workswithagents.dev/mcp-server/"
}
}
}Tools: bastion_fleet_health, bastion_trust_scores, bastion_agent_lookup, bastion_heartbeat, bastion_list_models, bastion_health
Complete 14-tool server with handoff protocol, Ed25519 identity verification, SLA validation, and compliance-as-code. Stdio transport — runs alongside your agent.
pip install wwa-mcpmcp_servers:
wwa:
command: python3
args:
- -m
- wwa_mcp.server| Tier | Tool | Server | Description |
|---|---|---|---|
| 1 | wwa_facts_query | Full | Query shared knowledge base — 68 facts |
| 1 | wwa_pitfalls_lookup | Full | Search 376 verified failure patterns |
| 1 | wwa_skills_search | Full | Find reusable agent workflows |
| 2 | wwa_handoff_create | Full | Structured agent-to-agent task handoff |
| 2 | wwa_handoff_verify | Full | Verify handoff completeness (0-10 score) |
| 2 | wwa_blueprint_search | Full | Search 15 CC BY 4.0 agent specs |
| 3 | wwa_trust_score_calculate | Full | Compute trust from 5 weighted signals |
| 3 | wwa_identity_verify | Full | Ed25519 signature verification |
| 3 | wwa_sla_validate | Full | SLA tier validation (3 tiers) |
| 3 | wwa_compliance_check | Full | Compliance-as-Code (11 operators) |
| — | bastion_fleet_health | Hosted | Fleet status — all 25 agents |
| — | bastion_trust_scores | Hosted | Trust score leaderboard |
| — | bastion_agent_lookup | Hosted | Find agent by ID |
| — | bastion_heartbeat | Hosted | Send heartbeat + trust signals |
| — | bastion_list_models | Hosted | Available LLM models |
| — | bastion_health | Hosted | Gateway health check |
| + 4 more: facts_categories, skills_categories, handoff_validate_schema, blueprint_validate | |||
25 autonomous agents report trust scores every 15 minutes. Scores computed from actual task success rates, pitfalls contributed, and uptime.