Sensitive vs. open
Mark a zone as sensitive and Synap will never send it to a remote provider. Open zones can use any model you've authorized.
MULTI-LLMBring your own LLM.
Bring your own LLM.
Claude, GPT, Gemini.
One mind, three model providers. Switch between them per task. Your knowledge stays yours; the model is just a substrate.
v1.4 · macOS · Windows · Linux
CLAUDEArchitecture review
sonnet-4.5 · activeGPTQuick brainstorm
gpt-4o · readyGEMINILong-context summary
2M tokens · ready3PROVIDERSClaude · GPT · Gemini
0LOCK-INbring your own keys
1MINDshared across providers
2 MINSETUPpaste a key, switch models
01 · ZONES, TASKS, MODELSThe right model
The right model
for the right task.
Synap routes each kind of work to the model that does it best — Claude for code, GPT for ideation, Gemini for long context.
Classified by intent
Architecture review, quick draft, long summary, code refactor — Synap recognizes the task and picks the right model defaults.
Pick per call
Override any default with a one-click model switcher. Your choice is remembered for the session, not enforced forever.
02 · COSTSPer-token costs,
Per-token costs,
visible per query.
See the cost-per-task before you run it. Synap meters every provider call, segments tokens by source (input · output · cache), and rolls up by zone — so you always know where the spend is going.
Claude · sonnet-4.512,480 in · 3,120 out
$0.087GPT · gpt-4o4,200 in · 980 out
$0.018Gemini · 1.5-pro64,300 in · 2,100 out
$0.041Claude · sonnet-4.58,100 in · 1,460 out
$0.062TOTAL · TODAY$0.208
VS LOCKED-IN AI
Pick the model. Not the cage.
Single-vendor AI tools tie you to one frontier. Synap stays neutral.
TODAY · SINGLE-VENDOR
ChatGPT-only or Claude-only
One vendor sees all your prompts
Stuck with their pricing changes
No way to A/B models per task
Vendor outage = your work stops
WITH SYNAP · MULTI-LLM
Best model per task
Three providers, one mind
Bring your own keys — no markup
Switch models per query, per zone
Outage on one? Swap to another in a click
03 · ROUTING & TRUSTYou decide
You decide
who sees what.
Per-zone, per-task model routing — with explicit trust tiers for every provider.
You decide which model
touches which file.
Pin sensitive zones to a local model. Route brainstorms to GPT. Send long-context jobs to Gemini. Synap enforces the rules per call — automatically.
Local models welcome.
Run Ollama, llama.cpp, or LM Studio? Synap talks to them like any other provider — and routes sensitive work there by default.
API key per provider, stored locally.
Keys live in your OS keychain. Synap never proxies them. Revoke and rotate without touching our infrastructure.
Substrate,
not synapse.
Free. macOS, Windows, Linux. Bring your own model.
Free · No credit card · No telemetry