The terminal coding assistant with real agent handoffs.
linkerassistant is an open-source CLI tool that integrates directly with your repository, letting you delegate complex refactoring and tasks to autonomous agents seamlessly.
✓ Workspace configured successfully.
Found relevant files: src/auth.js, src/middleware.js
Spawning agent session [id: auth-refactor-99a]
...
✓ Agent completed task. Diff ready for review.
_
Real repo. Real workflow. Real builders.
Built for environments where code quality and execution context actually matter.
Built around repository work, explicit permissions, and real agent handoffs.
LinkerAssistant isn't just a chatbot in your terminal. It's a structured workflow engine that delegates tasks to autonomous agents while keeping you firmly in control.
No hidden context
Agents only see what they explicitly need to. You control the scope of files passed to the model.
You control execution
Terminal commands are proposed by the agent, but never executed without your explicit Y/n approval.
Built for real codebases
Handles massive monorepos using advanced AST parsing and syntax-aware chunking.
Always interruptible
If the agent goes off track, send a SIGINT, correct the prompt, and resume seamlessly.
Every part of the workflow is a first-class surface.
Diff-based applies
Changes are staged as unified diffs, letting you review exactly what changed before committing.
Context awareness
Automatically pulls in imported files and type definitions relevant to the current task.
Permission boundaries
Strict sandboxing prevents agents from accessing environment variables or sensitive files.
History tracking
Every prompt, diff, and command execution is logged locally in `.linker/history`.
Local filesystem
Reads and writes directly to your disk. No remote syncing or cloud workspace required.
LSP integration
Leverages your local Language Server Protocol to detect compilation errors before applying.
Custom commands
Define custom scripts that the agent can invoke to run tests or linters automatically.
CI/CD pipelines
Run LinkerAssistant in headless mode to review PRs or automate boilerplate updates.
From trust to resume, in five steps.
The lifecycle of an autonomous task delegation.
TASK DELEGATION
Define the goal in natural language. LinkerAssistant parses the intent.
CONTEXT GATHERING
The agent reads relevant files and constructs a mental model of the codebase.
EXECUTION
Code is generated and shell commands are formulated to complete the task.
REVIEW & APPROVE
You inspect the generated diffs and approve terminal executions.
APPLY & RESUME
Changes are written to disk. The agent yields control back to you.
The terminal coding assistant, as a real workflow.
| Feature | LinkerAssistant | Standard AI Agents |
|---|---|---|
| Code edits | Unified diff review | Blindly overwrites files |
| Execution model | Prompt -> Plan -> Await Y/N | Auto-executes everything |
| Context boundaries | Explicit file inclusion | Sends entire directory |
| Environment | Your local terminal | Cloud container |
| Extensibility | Custom shell scripts | Hardcoded capabilities |
Everything important is already documented.
Questions that come up once you start using it for real work.
Can I use my own API keys?
Does it send code telemetry?
What models are supported?
How is context managed?
Can it run terminal commands?
Does it modify files directly?
Open the repo. Read the docs.
Run it locally.
The codebase is open source and available on GitHub. Stop copy-pasting code.
Start delegating workflows.