Mcp
Plith — Agent Infrastructure Suite
Hosted MCP exposing five agent-infrastructure products: DedupQ (task deduplication), BurnRate (cost prediction), QualityGate (output validation), GuardRail (behavioral governance), and PitfallDB (failure intelligence). One API key, ~14 MCP tools, credit-based pricing.
When to use Plith — Agent Infrastructure Suite
Choose if
You want a single hosted MCP that bundles agent-side infrastructure (dedup, cost prediction, output validation, governance, failure lookup) without wiring up five separate services. The free tier (1k credits/month) is enough to try all five products and decide which actually pull weight.
Avoid if
You need a specific best-in-class for any one of these jobs — DedupQ is not Redis; BurnRate is not OpenAI's usage API; QualityGate is not Guardrails.ai; GuardRail is not Lakera. Plith's leverage is the bundle, not depth. Also avoid until curator-fleet has verified tool semantics — the implementation is closed-source and the marketing copy is broad.
Risk Flags
- MEDIUM maturity Single-developer project (chicogonzales/plith); GitHub repo returns 404 so the implementation is a black box behind the hosted MCP at plith.ai/api/mcp. The marketing surface is polished but track record is limited; treat as experimental until curator-fleet eval confirms tool stability.
- LOW scope Each of the five products (DedupQ / BurnRate / QualityGate / GuardRail / PitfallDB) is broad in name but the actual tool depth isn't documented in the public registry entry. Probe `tools/list` before betting a workflow on a specific capability.
- LOW rate_limit Credit-based pricing; free tier is 1,000 credits/month with each tool call costing 1–3 credits, so free covers light development only.
Cost
Type: Freemium
Distribution
- MCP Registry
ai.plith/plith- Repository
- https://github.com/chicogonzales/plith