Bonito — The only neutral control plane above the hyperscalers
The structural position
Every other player in this category is structurally compromised:
- AWS Bedrock, Google Vertex AI, Azure AI Foundry are single-cloud by design. None of them can route to the others without killing their own lock-in business model. They will never be neutral.
- Kong AI Gateway, Apigee are API-gateway incumbents extending sideways into LLM. They reach the buyer through existing API-management contracts (real distribution advantage), but Apigee is Google-native, and Kong is a TCP/HTTP layer retrofitted to AI workloads.
- Portkey, Helicone, LiteLLM, Martian are developer-tier observability/routing proxies. Seed-stage funding, no enterprise governance surface, no agent layer, no KB layer, no compliance posture.
- Bonito is the only full-stack AI control plane that is structurally cloud-neutral. The moat is not a feature; it is a structural position that hyperscalers cannot replicate because it would break the business model they extract value from.
What is shipped in production today
- OpenAI-compatible gateway live across six AI providers: OpenAI, Anthropic, AWS Bedrock, Google Vertex AI, Azure AI, Groq. All six accessible via the same
bn- API key. - Cross-provider intelligent routing with five strategies (cost, latency, balanced, failover, A/B test) and automatic cross-region inference on Bedrock.
- Multi-provider failover on rate limits, timeouts, 5xx errors. Automatic retry on equivalent models.
- Immutable audit ledger across every model call, agent run, KB query, gateway request. Compliance teams answer their questions once.
- RAG knowledge bases on pgvector HNSW with 768-dim embeddings and VectorBoost 3.9-8x compression.
- Bonobot agents with visual canvas, default-deny tool policy, budget stops, rate limiting, SSRF protection, persistent agent memory (pgvector similarity), scheduled autonomous execution, approval queue with risk assessment.
- OpenAI-compatible chat completions, image generation (gpt-image-1, dall-e-3, dall-e-2), and video generation (Sora-2, Veo 2.0/3.0/3.1) on one key.
- Three token types — gateway keys (
bn-), personal access tokens (bp-), project tokens (bj-) — with per-tier caps and rate limits. - Multi-tenant org isolation, Vault-backed credential storage, AES-256-GCM at-rest encryption.
- SSO/SAML across Okta, Azure AD, Google Workspace, Custom SAML; RBAC; SOC-2 in flight; HIPAA, GDPR, ISO27001 governance checks across all three clouds.
- Origami: build agents, KBs, gateway keys, and provider connections by talking — a chat interface that orchestrates the platform instead of configuring it. New category, not a wrapper.
Production deployments today
Bonito is shipping in production across enterprise marketing-creative workflows, regulated-document processing, and customer-support agent flows. Customers run brand-asset generation pipelines, legal-doc triage, and CRM-attached support agents on the same gateway.
Pricing tiers
Free (3 providers, 25K req/mo, invite-only) · Builder $49/mo · Starter $199/mo (no procurement approval needed) · Growth $349/mo · Pro $999/mo · Enterprise from $6K/mo (typical band $6K-$20K) · Scale custom ($200K+/yr with dedicated infra and 99.99% SLA).
Founder
Founded 2025 by Shabari Shenoy — enterprise infrastructure background, prior to Bonito. New York based.
Verifiable references
Documentation (API reference, SDKs, integration guides) · Changelog (shipping cadence) · Pricing · Competitive matrix vs Portkey, LiteLLM, Helicone, Martian, Kong · Use cases · About · Contact (hello@trybonito.com)