alpnai-agent-kitSAFE
ALPNAI tools for authorized AI agents: deterministic cost, latency and quality analysis, reproducible examples, API/MCP integration and report delivery.
Overview
From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.
Français · Deutsch
ALPNAI gives authorized AI agents deterministic cost, latency and quality analyses. Compare variants on the same recorded tasks, then deliver a computed report to an owner-authorized project. Start by checking the public example before connecting your own data.
Have a CSV export? Import your own attempts before your first audit.
Start without an account: Trace Audit
Which recorded calls cost money without a confirmed result? Send up to 1,000 agent events to the free Trace Audit API. It returns computed costs, failed-task evidence and repeated-call candidates. No key, wallet or payment required.
python3 examples/trace_audit.py --input examples/trace-events.synthetic.json --report trace-result.json
Public MCP: https://alpnai.com/api/mcp/public · tools audit_agent_trace and get_trace_example, no authentication. Data contract and worked example · Try in your browser.
The main MCP endpoint and the paired comparison tools below retain their own access rules. Trace Audit does not change agents, store input events or promise savings.
Choose the analysis your agent needs
[Inspect the free example](https://alpnai.com/api/v1/performance-sample) · Connect through MCP · [Public G
8c2445a84880OBSERVED · 2026-10-12Connect
Built from this server's own package name, version and transport as found in its source — not copied from anyone's documentation, so it cannot drift against a page we do not control. Replace the environment placeholders with a token scoped to the least it needs.
claude mcp add alpnai-buyer-client-draft --env ALPNAI_AGENT_KEY=${ALPNAI_AGENT_KEY} --env ALPNAI_MONITOR_TOKEN=${ALPNAI_MONITOR_TOKEN} --env ALPNAI_SITES_BYPASS=${ALPNAI_SITES_BYPASS} -- npx -y [email protected]{
"mcpServers": {
"alpnai-buyer-client-draft": {
"command": "npx",
"args": [
"-y",
"[email protected]"
],
"env": {
"ALPNAI_AGENT_KEY": "${ALPNAI_AGENT_KEY}",
"ALPNAI_MONITOR_TOKEN": "${ALPNAI_MONITOR_TOKEN}",
"ALPNAI_SITES_BYPASS": "${ALPNAI_SITES_BYPASS}"
}
}
}
}Trust audit
SAFEgrade B · trust 87/100 Nothing in the source contradicts what it says it does. Grade A is reserved for packages that have also passed the behavioural sandbox.
| Layer | What it checks | Result |
|---|---|---|
| L0 | Provenance & inventory | PASS |
| L1 | Static analysis of the code | WARN |
| L2 | Instruction surface (what it tells the agent) | PASS |
| L3 | Class-specific surface | PASS |
| L4 | Behavioural (sandbox) | SKIPPED |
What the source does
- Filesystem
- none-observed
- Network
- declared (1 observation(s))
- Shell
- none-observed
- Dependencies
- pinned
- Secrets in source
- found
Findings (11)
TOKEN = "synthetic_monitor_secret"
server=ThreadingHTTPServer(('127.0.0.1',0),Handler);base=f'http://127.0.0.1:{server.server_port}'for base in ['https://attacker.example','https://alpnai.com.attacker.example','https://[email protected]','https://alpnai.com/path','http://alpnai.com','https://alpnai.com:444','http://127.0.0.1:9999']:
"--base-url", f"http://127.0.0.1:{server.server_port}"],test('Standalone CLI succeeds with fixtures on loopback and GET requests only',async()=>{const calls=[],server=createServer((req,res)=>{calls.push({method:req.method,path:req.url,authorization:req.heawith self.assertRaises(client.AuditClientError): client.audit(base=f"http://127.0.0.1:{server.server_port}", key=KEY, raw=json.dumps(INPUT).encode())The [new payment-monitoring update](payment-monitoring.md) was published on 14 September 2026 at 11:21 UTC. An authenticated manual pass at 11:24 UTC was then observed in the private console. It keeps
`audit_agent_costs` accepts the same `{runs, config?}` input as `POST /api/v1/spend-proof`. Send an active `Authorization: Bearer <pilot-test-key>` and `Mcp-Name: audit_agent_costs`. It consumes no pu**Python kit:** from the root of the [ALPNAI agent kit](https://github.com/fredericmagnathy-ops/alpnai-agent-kit), supply an active key through your process secret manager as `ALPNAI_AGENT_KEY`, then
Secrets are scoped to the individual probe step. The workflow token has `contents: read`; checkout does not persist credentials. The Python client refuses HTTP redirects, bounds each network operation
Gates applied: no_behavioural_pass.
8c2445a84880full audit observations/trust-audit/mcp-server/fredericmagnathy-ops__alpnai-agent-kit.json · Report an issue / request a re-scanAudit history
Every audit this server has had. A grade with a past is a grade somebody is still checking.
| Date | Source | Verdict | Grade | Score | Change |
|---|---|---|---|---|---|
| 2026-10-12 | 8c2445a84880 | SAFE | B | 87 | first audit |
Questions
What is the alpnai-agent-kit MCP server?
ALPNAI tools for authorized AI agents: deterministic cost, latency and quality analysis, reproducible examples, API/MCP integration and report delivery.
Is alpnai-agent-kit safe to connect to an agent?
The audit found nothing in the source that contradicts what it says it does, and graded it B (87/100). Grade A is held back for packages that have also passed a sandboxed behavioural run, which is why a clean server reads B.
What credentials does alpnai-agent-kit need?
It reads ALPNAI_AGENT_KEY, ALPNAI_MONITOR_TOKEN and ALPNAI_SITES_BYPASS from the environment. Give it a token scoped to the least it needs — an agent that can be talked into calling a tool can be talked into calling it with your credentials.
How does alpnai-agent-kit run?
It speaks streamable-http, so it runs as a service you connect to over the network. It is published on npm as alpnai-buyer-client-draft at 0.0.0-draft.
How current is this page?
The grade is for one exact copy of the source (8c2445a84880), read on 2026-10-12. The repository is watched and re-audited when it changes.