General
Per-agent model assignments. Each agent can use any provider and model independently.
Agent Configuration
AWS Bedrock
Configure your AWS region, authentication, and credentials. Credentials are stored in VS Code SecretStorage.
Connection
Credentials — Stored Securely
Leave blank to use the default regional endpoint.
Available Models
Fetch live from your AWS account — populates the model dropdown.
⚠ Models shown are available in your region but must be enabled in AWS Console → Bedrock → Model Access before use. If you see "Operation not allowed", enable the provider there first.
Add Model Manually
For newer inference profiles not returned by the list API.
Interpreter DB
Auto-learned response formats for unknown models. Stored at ~/.jinn-code/interpreters.json.
Learned Models
Loading…
Kernel Scheduler
Routes each AI task to a specific model. Changes take effect on the next message.
Preset
⚡ Performance
All tasks use your Thinker model. Maximum capability.
⚖ Balanced
Heavy tasks use Thinker/Coder, fast tasks use Planner/Reviewer. Best value.
⚙ Custom
Define each task's model manually.
Active Dispatch Table
| Task | Role | Provider | Model ID |
|---|
Local Models
Connect to Ollama, LM Studio, or any OpenAI-compatible local server.
Local Server
Select from dropdown after fetching, or type manually.
Common Models
llama3.1:8bOllama
llama3.1:70bOllama
deepseek-coder-v2:16bOllama
qwen2.5-coder:7bOllama
mistral-nemo:12bOllama
MCP Servers
Model Context Protocol servers extend agents with custom tools. No settings.json editing required.
Configured Servers