Loading ToolBench
Hosted MCP server providing lossless compression, pay-per-suite pricing, x402 agentic commerce, and product catalog tools
Slid Phi Labs MCP exposes 17 tools via HTTP with basic schemas and descriptions. However, critical gaps undermine production readiness: (1) Most tool descriptions lack actionable context, they read as feature summaries rather than LLM-optimized guidance. E.g., 'spl_discover' has a 73-char description with no explanation of when/why to call it. (2) Parameter descriptions are sparse or missing entirely. 'spl_quote' accepts 'product' and 'dataClass' enums but provides no guidance on valid values or dependencies. (3) Output schemas are not documented, LLMs cannot predict response structure for chaining. (4) Error handling is absent, no recovery guidance, no categorization of retryable vs. fatal errors. (5) Tool naming is inconsistent: 'spl_pps_x402_buy' vs 'spl_pps_x402_submit' are similar but distinct; LLMs will conflate them. (6) Security: 'spl_pps_x402_buy' and 'spl_pps_x402_submit' accept payment headers and email as parameters, credentials in params risk logging leaks. (7) Composition: 'spl_quote' and 'spl_pps_quote' appear to duplicate functionality; unclear which to use. Average tool score: 42/100.
CuNi Bank: paste N, get X. Ingest → emit → prove, or refuse. v1 from Python. Args: source, from (py|cuni), to (js|go|py|ts).
List standing Slid Phi Labs seats agents can buy via x402 (Chamber, PCC, Rider, Lab Pass, consulting)
Hosted lossless compression. Every dual-licensed pathway on the lab machine. Args: data_b64. Returns packed_b64. First 2 GB/month free, then 8¢/GB.
Restore a file packed by spl_compress. Args: packed_b64. Returns raw_b64.
Agent discovery endpoint for agentic commerce rails (Stripe human + x402 agent)
One lab account for every Slid Phi product. POST https://www.slidphilabs.com/api/auth {action:signup|login,email,password,name}. Then open CuNi, Rider, Chamber with the token. Humans: /signup /login /account.
Output schemas not documented. LLMs cannot predict response structure, breaking tool chaining and forcing discovery detours.
Parameter descriptions missing or incomplete. 'product' and 'dataClass' enums lack guidance on valid values and dependencies. LLMs will guess or hallucinate.
Credentials and payment headers exposed as tool parameters (paymentHeader, email in spl_pps_x402_buy, spl_pps_x402_submit). Secrets in params leak into logs and traces.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 55 | 2026-07-28+ | v2 |
Checkout for suite quote. Under 2 GB/month returns free_showcase (no Stripe). Over free returns Stripe Checkout URL.
Classify a base64 sample (first few KB) into a data class and recommended tool. No proprietary encode.
SPL Pay Per Suite overview: First 2 GB each month free · then 8¢/GB · $1 card minimum. Stripe (humans) + x402 (agents). Seats: Chamber $9/$99 · PCC $9/$49/$490 · Rider $79/$790 · Lab Pass $668 · consulting $250.
Instant quote: free first 2 GB/month ($0), then 8¢/GB · $1 card minimum. Inputs: product, dataClass, op, bytes.
Submit a Pay Per Suite job (free under 2 GB/month after free_showcase, or after paid). Lab runs best tool and emails results. Requires email.
Buy a standing license product via x402. Without paymentHeader → returns 402 requirements. With paymentHeader (or devBypass) → order_id + access_url. Not for suite jobs — use spl_pps_x402_submit.
List standing Slid Phi Labs seats agents can buy via x402 (Chamber, PCC, Rider, Lab Pass, consulting). Suite jobs are separate.
Agentic commerce discovery: standing product catalog + freemium suite jobs (free under 2 GB/month), headers, flow. Stripe remains for humans.
Probe POST /api/x402-suite. Free under 2 GB/month (no pay). Over free → 402 accepts[] for usage (8¢/GB · $1 card minimum).
Submit freemium suite job via x402. Under 2 GB/month free (no paymentHeader). Over free: pass paymentHeader after paying, or devBypass for staging.
Instant quote: free first 2 GB/month ($0), then 8¢/GB · $1 card minimum. Inputs: product, dataClass, op, bytes.
Duplicate/overlapping tools: spl_quote vs spl_pps_quote, spl_catalog vs spl_pps_x402_catalog. LLMs cannot disambiguate; composition breaks.
No error handling guidance. Tools lack recovery hints, error categorization (retryable vs fatal), or actionable error messages. Agents cannot self-correct.