Comparison
BlockRun vs Portkey
Both sit in front of multiple LLM providers. Portkey's focus is enterprise AI ops (observability, caching, guardrails, governance) on a subscription. BlockRun's focus is autonomous agents paying per-call in USDC with no account.
Side-by-side
| Feature | BlockRun | Portkey |
|---|---|---|
| Billing | Pay-per-call USDC; no minimum | Tiered subscription + per-call overages |
| Auth | Wallet signature per call | API key |
| Free tier | Yes, no card | Free dev tier (rate-limited) |
| Auto-fallback | Built-in cross-provider chain | Yes (configurable) |
| Observability | Basic metrics + on-chain receipts | Deep — traces, evals, prompt versioning, A/B tests |
| Caching | No (each call settles independently) | Yes, semantic + simple |
| Guardrails (PII, content filters) | No | Yes |
| Self-hosted option | No (Cloud Run gateway) | Yes (enterprise) |
| BlockRun MCP | Yes (18 tools) | No |
| Settlement | On-chain (Base/Solana) | Off-chain billing |
When to choose BlockRun
- You're shipping autonomous AI agents that pay for their own inference.
- You don't want to manage prepaid credits, subscriptions, or seat counts.
- You need on-chain payment receipts (auditable, agent-to-agent).
- You want zero-signup “just try it” access for users.
- You're using Claude Code, Cursor, or other MCP clients.
When to choose Portkey
- You're running production AI workloads at an enterprise and need deep observability, prompt versioning, and evals.
- You need built-in guardrails (PII redaction, content filtering) at the gateway layer.
- You need to self-host the gateway inside your VPC.
- You want semantic caching to cut repeated-query costs.
- You're fine with a subscription billing model and a single API key.
The bottom line
Portkey is a deeper enterprise AI ops platform — caching, evals, guardrails, prompt management, observability. BlockRun is a payment-rail-first gateway — per-call USDC, no signup, agent-native MCP. They're complementary: an enterprise can route Portkey's outbound traffic through BlockRun for on-chain settlement. Most teams pick based on whether the unit of value is “production observability” (Portkey) or “autonomous agent spend” (BlockRun).
If something on this page is wrong, email hello@blockrun.ai.