Tools

Oh My OpenAgent gives your agents world-class tools. LSP, AST-Grep, Tmux, MCP — actually integrated, not duct-taped together.

Hashline (Hash-Anchored Edit Tool)

The harness problem is real. Most agent failures aren’t the model — it’s the edit tool.

“None of these tools give the model a stable, verifiable identifier for the lines it wants to change… They all rely on the model reproducing content it already saw. When it can’t - and it often can’t - the user blames the model.”

— Can Bölük, The Harness Problem

Inspired by oh-my-pi, Hashline tags every line with a content hash:

11#VK| function hello() {
22#XJ|   return "world";
33#MB| }

The agent edits by referencing those tags. If the file changed since the last read, the hash won’t match and the edit is rejected before corruption. No whitespace reproduction. No stale-line errors.

Impact: Grok Code Fast 1 success rate went from 6.7% → 68.3% — just from changing the edit tool.

IntentGate

Analyzes true user intent before classifying or acting. No more literal misinterpretations.

When you type a request, IntentGate:

Parses what you actually mean (not just what you said)
Classifies the task type
Routes to the appropriate agent

This prevents agents from taking instructions too literally and producing unexpected results.

LSP Tools

IDE precision for every agent:

Tool	What it does
`lsp_rename`	Safe cross-file symbol renaming
`lsp_goto_definition`	Jump to symbol definition
`lsp_find_references`	Find all references to a symbol
`lsp_diagnostics`	Pre-build diagnostic checks

These give agents the same code intelligence that your IDE provides — type-safe refactoring, accurate navigation, and real-time error detection.

AST-Grep

Pattern-aware code search and rewriting across 25 languages.

Unlike text-based search, AST-Grep understands code structure:

Search by syntax pattern, not string matching
Rewrite code while preserving structure
Works across TypeScript, Python, Go, Rust, Java, and 20+ more languages

Example patterns:

Find all React components using useState
Find all async functions without error handling
Batch-replace deprecated API calls

Tmux Integration

Full interactive terminal. REPLs, debuggers, TUI apps. Your agent stays in session.

What Tmux enables:

Run long-running processes (dev servers, test watchers)
Interact with REPLs (Node.js, Python, IRB)
Use debuggers (gdb, pdb, node —inspect)
Run TUI applications (vim, htop, lazygit)
Agent maintains session state across interactions

Skill-Embedded MCPs

MCP servers eat your context budget. Oh My OpenAgent fixed that.

Skills bring their own MCP servers:

Spin up on-demand — only when the skill is active
Scoped to task — only relevant tools exposed
Gone when done — context window stays clean

Built-in Skills

Skill	What it does
`playwright`	Browser automation — screenshots, E2E tests, web scraping
`git-master`	Atomic commits, rebase surgery, clean git history
`frontend-ui-ux`	Design-first UI implementation

Custom Skills

Add your own skills:

Project-level: .opencode/skills/*/SKILL.md
User-level: ~/.config/opencode/skills/*/SKILL.md

Each skill includes:

Domain-tuned system instructions
Embedded MCP server definition
Scoped permissions

Built-in MCPs

Three MCP servers are always on:

MCP	Based On	What it does
`websearch`	Exa	Web search — find relevant code, docs, solutions
`context7`	Context7	Official documentation lookup — fetch library docs into context
`grep_app`	Grep.app	GitHub code search — search across public repositories

These are configured by default and require no setup. They provide agents with real-time access to web information, documentation, and code search.

Next Steps

Agent System — How agents use these tools
Features — Ralph Loop, Todo Enforcer, and other features
Configuration — Configure MCPs, LSP, and tool behavior