AI gateways

Bonito vs TrueFoundry AI Gateway + Agent Gateway + TrueForge

Gateway plus agent harness vs. gateway plus managed agents

Launched Agent Gateway (June 2026) and the MIT-licensed TrueForge agent harness (August 2026)

TrueFoundry is an enterprise AI gateway (LLM, MCP and agent traffic) that runs in your VPC, on-prem or air-gapped, and now ships TrueForge, an open-source agent harness with approval checkpoints and sessions. It is the closest product to Bonito. The differences are in how agents are built and run: Bonito adds a visual canvas and chat builder, schedules, integrations, managed knowledge bases and data residency pinning, all from one bonito.yaml.

TrueFoundry pricing

Developer free (3 users); Pro $25 per user per month; Enterprise custom (SSO, audit logs, VPC and on-prem). Agent Gateway and harness: custom.

Model costs

No markup. Model providers bill you directly on your own accounts.

What TrueFoundry does well

  • Routing by cost, latency and geography with fallback chains over 250+ models, including self-hosted ones
  • Deploys in your VPC, on-prem or air-gapped
  • Agent Gateway with an agent registry, per-agent budgets and end-to-end traces across model and tool calls
  • TrueForge (MIT): subagents, human approval checkpoints and persistent sessions
  • GitOps with tfy apply on YAML specs, plus semantic caching scoped per user

Where Bonito goes further

  • A visual agent canvas and a chat builder (Origami); no visual builder was found in TrueFoundry's docs
  • Cron schedules with timezone and delivery to webhook, email or Slack
  • Managed knowledge bases with upload, embedding, search and a BYO bucket; TrueFoundry's Cognita is a RAG framework you deploy
  • Data residency pinning to CA, US or EU as one org setting that also constrains failover
  • Native integrations (Slack, Stripe, Google Workspace, HubSpot, Notion, Twilio) bound to agents as tools, including two-way Slack chat

Where TrueFoundry is ahead of Bonito

  • Self-hosting, including air-gapped installs, and an MIT-licensed harness
  • Semantic caching; Bonito has none
  • Self-hosted model serving alongside the gateway

Capability by capability

CapabilityBonitoTrueFoundry
Agents
Build and run agentsYesYescode-first TrueForge harness; no visual builder found
Approval queue and scheduled runsYesPartialapproval checkpoints in TrueForge; no cron schedules found
Dollar budgets per agentYesYes
MCP supportYesYes
Routing
OpenAI-compatible gatewayYesYes
Automatic failover across providersYesYes
Cheaper models for simple work (cost routing)YesYes
Bedrock, Vertex, Azure and direct providers in one placeYesPartial250+ models; which clouds is not itemized in their docs
Semantic cachingNoYes
Governance
SSO, RBAC and audit logYesYes
Data residency pinningYesYesgeo-aware routing plus VPC or on-prem
Observability and knowledge
Agent traces with cost per request and agentYesYes
Knowledge bases (RAG)YesPartialCognita, an open-source RAG framework you deploy
How you run it
Whole stack from one config fileYesYestfy apply on YAML specs
Self-hosted or open sourcePartialYesgateway self-hosts on Enterprise; TrueForge is MIT
No markup on inferenceYesYes

Checked against TrueFoundry's own documentation on 2026-09-27. Products change quickly; if something here is out of date, tell us at hello@trybonito.com and we will fix it.

Sources

See Bonito on your own stack

Start free, or get a report on where agents fit in your company first.