Get This Tool
Switchboard
Pricing
- Model
- Free
Summary
Running Codex for code generation, Claude Code for reasoning, and Ollama for everything else means every prompt either leaks to a premium API you didn't intend or burns quota you were trying to save — there's no middle layer making that call for you. Switchboard is that middle layer.
Switchboard sits between your terminal and your CLI coding agents, inspecting each prompt and routing it to the right backend based on rules you define — local Ollama for the routine stuff, Codex for code, Claude Code when reasoning depth is worth the cost. The vendor's own benchmark reports 62% fewer premium-agent calls with a quality drop from 4.6 to 4.1 on a five-point scale — acceptable for most teams, not for teams where output consistency is non-negotiable. Sensitive prompts stay on-device by design, with zero benchmark leaks observed in testing. Routing decisions are logged with metadata-only telemetry, so you can audit what went where without storing prompt content. It ships with a Dockerfile and pyproject.toml, which means self-hosting is the deployment model — there is no hosted service.
Bottom line: Pick Switchboard if you are already paying for premium CLI agents and want a rules layer that keeps sensitive prompts local and trims your bill; skip it if your team needs a web UI, an API surface, or quality parity with always-premium routing.
Community Performance Report Card
No community ratings yet. Be the first to rate this tool!
Community Benchmarks Community
Sign in to submit a benchmarkNo community benchmarks yet. Be the first to share a real-world data point.
Pros
Sign in to edit- Deterministic on-device prompt filtering before any network call, which means sensitive prompts cannot reach a cloud API even if your routing rules are misconfigured downstream.
- Unified session across Ollama, Codex, and Claude Code, so context you built up in one backend is not lost when the router hands off to another — avoiding the manual copy-paste loop developers run today.
- Metadata-only telemetry for routing decisions, which means you get an audit trail of what went where without storing prompt content that could become a compliance liability.
- Vendor-reported 62% reduction in premium-agent calls at a 4.1 vs 4.6 quality score, so teams can set a quality threshold and let the router handle cost optimization rather than manually triaging every prompt.
- MIT-licensed with Docker and Python packaging included, which means you can fork, modify, and run it in an air-gapped environment without negotiating a commercial license.
Cons
Sign in to edit- No API surface exists, so any service or script that needs to call the router programmatically cannot — teams that want to integrate Switchboard into a CI pipeline or a web app hit a dead end and either fork the project to add an API layer or replace it with a router that exposes one.
- The quality delta between routed and always-premium output is real: the vendor's own benchmark puts routed quality at 4.1 versus 4.6 for always-premium. Teams where output feeds customer-facing copy or code review tooling will find that gap shows up in production reviews, and they end up tuning rules until the router sends most prompts to premium anyway — at which point the cost savings disappear.
- With 2 GitHub stars and no open issues at time of curation, the community is effectively zero. When a routing edge case surfaces in production — and it will — there is no forum thread, no maintainer response history, and no third-party fix to pull. Teams with low tolerance for debugging undocumented behavior switch to a more established local proxy with an active issue tracker.
Community Reviews
Sign in to write a reviewNo reviews yet. Be the first to share your experience.
About
- API Available
- No
- Self-Hosted
- Yes
- Last Updated
- 2026-07-01T02:23:21.830Z
Best For
Who it's for
- Developers using multiple CLI coding agents
- Users requiring strict on-device prompt handling
- Local LLM workflows needing seamless backend switching
What it does well
- Route coding prompts between local and premium backends
- Enforce deterministic privacy rules on sensitive prompts
- Maintain context across Ollama, Codex, and Claude Code sessions
- Reduce premium CLI agent quota usage with quality retention
- Log routing decisions with metadata-only telemetry
Integrations
Discussion Community
Sign in to commentNo discussion yet. Sign in to start the conversation.
Compare Switchboard
Spotted incorrect or missing data? Join our community of contributors.
Sign Up to ContributeCommunity Notes & Tips Community
Sign in to contributeBe the first to contribute. General notes, observations, gotchas, and tips from people who use this tool day-to-day.
Frequently Asked Questions
- Is Switchboard free?
- Yes — Switchboard is fully free to use. There is no paid tier.
- Is Switchboard open source?
- Yes. Switchboard is open source.
- Can I self-host Switchboard?
- Yes. Switchboard supports self-hosting on your own infrastructure.
Hours Saved & ROI Stories Community
Sign in to contributeBe the first to contribute. Concrete time/cost savings, with context. e.g. "Cut my code review backlog from 4h to 45m per week."
Curated lists that include this category
Switchboard is an open-source, self-hosted router that wraps Codex, Claude Code, and Ollama behind a single session interface. When a prompt arrives at the CLI, Switchboard applies your configured routing rules — sending code-generation requests to Codex, reasoning tasks to Claude Code, and everything else to a local Ollama instance. The session stays continuous across all three backends, so context does not reset when the backend switches. No hosted service exists; you run it yourself via Docker or Python.
The defining feature is on-device privacy enforcement. The docs describe deterministic rules that classify prompts before any network call is made, so prompts matching your sensitive-data patterns never leave the machine. The vendor states zero benchmark leaks were observed during evaluation. That guarantee comes from the local-first architecture rather than from post-hoc filtering, which means it holds even if a cloud API is unavailable.
This tool fits developers who already have CLI agent subscriptions and want a cost-control layer without rebuilding their workflow. It breaks down when teams need a no-code configuration surface, a REST API other services can call, or a quality floor that matches always-premium routing — the reported quality delta is real, and teams with customer-facing output pipelines will feel it. There is no community fork ecosystem yet; the GitHub page shows 2 stars and 0 open issues, so production edge cases land on you to debug.
Deployment uses the included Dockerfile and docker-compose.yml. Configuration lives in the config directory, and an .env.example file covers credential wiring for each backend. The project is MIT-licensed with a CITATION.cff, suggesting the author targets both engineering and research audiences.