How it compares
| SoulForge | Claude Code | Copilot CLI | Aider | |
|---|---|---|---|---|
| Editor | Embedded Neovim (your config) | No editor | No editor | No editor |
| Code graph | SQLite graph — PageRank, blast radius, cochange, clone detection, FTS5 | None (file reads + grep) | None | Tree-sitter repo map + PageRank |
| Code intelligence | 4-tier fallback: LSP, ts-morph, tree-sitter, regex. Dual LSP. 12 operations. 33 languages | LSP via plugin (5 ops, no rename) | No agent LSP access | Tree-sitter AST |
| Compound tools | rename_symbol, move_symbol, refactor — compiler-guaranteed, atomic | Rename via LSP | — | — |
| Semantic context | AST summaries for top 500 symbols, dispatch auto-enrichment | — | — | Tree-sitter tag summaries |
| Multi-agent | Parallel dispatch (8 agents, 3 concurrent, shared cache, edit ownership) | Subagents + Agent Teams | Subagents + Fleet | Single agent |
| Providers | 10 (Anthropic, OpenAI, Google, xAI, Ollama, +5) | Anthropic only | Multi-model | 100+ LLMs |
| Task routing | Per-task model assignment (plan, code, explore, search, trivial, cleanup, compact) | Single model | Single model | Single model |
| Context management | 2-layer: per-step pruning + V1/V2 compaction | Auto-compaction | Context window management | — |
| License | BSL 1.1 (source-available) | Proprietary | Proprietary | Apache 2.0 |
Highlights
Embedded Neovim
Your Neovim with LazyVim, LSP via Mason, treesitter highlighting. The AI edits through the same editor you use.
Multi-agent dispatch
Parallel explore, code, and web search agents with shared file cache and edit coordination.
Graph-powered repo map
SQLite codebase graph with PageRank, cochange analysis, blast radius, and clone detection.
4-tier code intelligence
LSP, ts-morph, tree-sitter, regex fallback chain across 33 languages with dual LSP backend.
Compound tools
rename_symbol, move_symbol, refactor, project do the complete job in one call.Task router
Assign models per task: Opus for planning, Sonnet for coding, Haiku for search.
Context management
Rolling tool-result pruning per step, plus V1/V2 compaction on threshold.
User steering
Type messages while the agent works — they queue and inject into the next step.
10 providers
Anthropic, OpenAI, Google, xAI, Ollama, OpenRouter, and custom OpenAI-compatible APIs.
Cross-tab coordination
Up to 5 concurrent tabs with advisory file claims and git blocking during dispatch.
Project toolchain
Auto-detects lint, typecheck, test, and build across 21 ecosystems with pre-commit enforcement.
Skills and approval gates
Installable skills for domain-specific capabilities. Destructive actions individually prompted.
Tool suite
SoulForge ships 31 tools organized by capability:Code intelligence
| Tool | What it does |
|---|---|
read_code | Extract function/class/type by name (LSP-powered) |
navigate | Definition, references, call hierarchy, implementations |
analyze | File diagnostics, unused symbols, complexity |
rename_symbol | Compiler-guaranteed rename across all files |
move_symbol | Move to another file + update all importers |
refactor | Extract function/variable, organize imports |
Codebase analysis (zero LLM cost)
| Tool | What it does |
|---|---|
soul_grep | Count-mode ripgrep with repo map intercept |
soul_find | Fuzzy file/symbol search, PageRank-ranked, signatures included |
soul_analyze | Identifier frequency, unused exports, file profiles, top files, packages, symbols by kind |
soul_impact | Dependency graph — dependents, cochanges, blast radius |
Project management
| Tool | What it does |
|---|---|
project | Auto-detected lint, format, test, build, typecheck across 21 ecosystems |
project(list) | Discover monorepo packages with per-package capabilities |
dispatch | Parallel multi-agent execution (up to 8 agents, 3 concurrent) |
git | Structured git operations with auto co-author tracking |
LLM providers
| Provider | Models | Setup |
|---|---|---|
| Anthropic | Claude Opus 4.6, Sonnet 4, Haiku 3.5 | ANTHROPIC_API_KEY |
| OpenAI | GPT-4.5, o3, o4-mini | OPENAI_API_KEY |
| Gemini 2.5 Pro/Flash | GOOGLE_GENERATIVE_AI_API_KEY | |
| xAI | Grok 3 | XAI_API_KEY |
| Ollama | Any local model | Auto-detected |
| OpenRouter | 200+ models | OPENROUTER_API_KEY |
| LLM Gateway | Multi-model gateway | LLM_GATEWAY_API_KEY |
| Vercel AI Gateway | Unified gateway for 15+ providers | AI_GATEWAY_API_KEY |
| Proxy | Local proxy with auto-lifecycle | PROXY_API_KEY |
| Custom | Any OpenAI-compatible API | Any env var |
Read the full story on why we built SoulForge.