# dirac-run/dirac

Coding Agent singularly focused efficiency and context curation. Reduces API costs by 50-80% vs other agent AND improves the code quality at the same time. Uses Hash Anchored edits, massively parallel operations, AST manipulation and many many other optimizations.  https://dirac.run/

Repository: https://github.com/dirac-run/dirac
Canonical: https://ross.abutalabs.com/products/dirac
Language: TypeScript
License: Apache-2.0
License Family: permissive
Last push: 2026-08-26T17:44:39+00:00

## Health v2 (maintenance only)
Score: 81/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 99, release rhythm 99, longevity 10
- inputs: {"age_days": 150, "days_push": 7, "days_rel": 7, "gap_med": 3, "n_releases_24m": 6}
- flags: young
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1465, forks 83 (observed 2026-08-28T04:04:48.673109+00:00)

## What it is
Dirac is an open-source AI coding agent focused on token efficiency and context curation, claiming 50-80% API cost reduction versus other agents while improving code quality. It uses hash/line-anchored parallel edits, AST manipulation, and other optimizations, and topped the Terminal-Bench 2 leaderboard.

## Use cases
- reduce LLM API costs for AI coding agents
- automated code refactoring with an AI agent
- terminal-based coding assistant with curated context
- run coding agent benchmarks like Terminal-Bench
- make large multi-file code changes in a single task
- open-source alternative to Cline or Roo

## When to choose
- you want a cost-efficient terminal coding agent with strong benchmark results
- you work on large refactoring tasks where context length degrades model accuracy
- you prefer an open-source agent without MCP dependencies

## When to avoid
- you need MCP tool integrations or a GUI/IDE plugin experience
- you rely on features specific to other agent ecosystems
- the project's cost claims need independent verification for your workload

## Facets
- artifact type: cli-tool
- maturity: active
- function: agent-framework, developer-tools, cli, llm-inference
- domain: developer-tools, artificial-intelligence
- platform: cli, cross-platform
- tags: coding-agent, token-efficiency, context-curation, ast-manipulation, terminal-bench, cost-optimization, ai-agents, command-line, nodejs

## Member repositories
- dirac-run/dirac (main) score 81

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:48.673109+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T04:35:05.492794+00:00, confidence not recorded.
  - readme: https://github.com/dirac-run/dirac (fetched 2026-08-28T04:04:48.673109+00:00, sha d5fcc863d1fa)
- Data as of 2026-08-30T08:39:29.467469+00:00.