AgentIndex icon
AgentIndex
ToolsMCP ServersHOTCategoriesTrendingNewBest ListsCompareGuides
Submit Tool
ToolsMCP ServersHOTCategoriesTrendingNewBest ListsCompareGuides
Home/
Code Assistant/
ATLAS
ATLAS logo

ATLAS

Active·★ 2.1k·AGPL-3.0·Updated 2026-11-16

A.T.L.A.S. ('Adaptive Test-time Learning and Autonomous Specialization') by Inferstep is an open-source, local-first terminal coding agent (AGPL-3.0) designed to bring frontier-grade reasoning to compact open-weight models running on consumer hardware.

Addressing the reality that raw 7B–14B models struggle with complex multi-file software engineering tasks when relying on single-shot completions, ATLAS surrounds the model with an intelligent system scaffolding. It implements test-time compute scaling: dynamically decomposing goals into hierarchical plans, generating multiple candidate code solutions, ranking candidates via quality scoring, compiling and executing code inside isolated sandboxes, and automatically repairing test failures. Compatible with GGUF models via llama.cpp, ATLAS runs locally across Apple Silicon, NVIDIA, AMD, and CPUs with zero cloud API fees.

For privacy-conscious developers and offline engineering teams, ATLAS unlocks remarkable code generation capabilities without external token billing. However, allocating substantial test-time compute (generating multiple candidates and running local test suites) multiplies local inference duration, smaller compact models still face reasoning limits on high-level architecture, and the codebase carries copyleft AGPL-3.0 licensing.

This write-up is based on the project's README and public repository data. We have not hand-tested the tool. Data as of 2026-11-12.

License
AGPL-3.0
Latest release
v3.1.7 · 2026-10-25
Last commit
2026-11-12
Contributors
12
Forks
201
Open issues and PRs
147
First published
2026-02-01

Source: GitHub, npm and PyPI public APIs.

#agent-framework#ai-agent#ai-harness#coding-agent#developer-tools#llm#local-first#local-llm
↗ Visit site★ GitHub
01

Features

01Planning, candidate selection, verification and repair around the model
02Sandboxed compile and test before accepting code
03Runs GGUF models on NVIDIA, AMD, Apple Silicon, Vulkan or CPU
04No hosted model or third-party API key required
05Terminal UI
02

Why choose it

+Empowers compact 7B-14B open models to achieve frontier-like coding accuracy via test-time compute scaffolding.
+Completely local and offline: runs GGUF models on Apple Silicon, NVIDIA, AMD, and CPU with zero cloud fees.
+Integrated execution sandbox compiles, tests, and iteratively repairs code before presenting changes to the user.
+Adaptive compute allocation: simple edits take fast paths while difficult tasks trigger multi-candidate reasoning.
03

Trade-offs

!Test-time multi-candidate generation and iterative local test execution noticeably increase overall completion latency.
!Smaller 7B-14B models can still hit cognitive reasoning ceilings when tasked with massive architectural redesigns.
!Licensed under AGPL-3.0, which imposes strong copyleft requirements on commercial software integrations.
!Heavy local inference loops can maximize CPU/GPU utilization, impacting battery life on laptops.
04

Maintenance check

ActiveChecked: 2026-11-12
Last commit
2026-10-24
Commits, last 90 days
55
Releases, last 90 days
4
Latest release
v3.1.7 (2026-10-25)
Open issues
147
Recent issues with at least one comment
0/1
Commits per week, last 12 weeks
Commits per week, last 12 weeks: 0, 0, 0, 0, 1, 18, 10, 0, 26, 0, 0, 0

Oldest week on the left; the latest four weeks are orange. Source: GitHub commit activity.

Status rule: archived on GitHub = Archived; last commit within 90 days = Active; 91 to 365 days = Slowing; older = Stale. All figures come from the GitHub API on the date shown.

Most discussed open issues

  • #35 JetBrains plugin (the VS Code extension is on dev, for 3.2.0)16 comments · 2026-04-13
  • #245 CI: run each service's Python tests on the Python its image uses6 comments · 2026-09-28
  • #238 Held-out evaluation suite: frozen tasks and graders with paired controls6 comments · 2026-09-28
  • #215 Content loops are detected late and use more than half of a session6 comments · 2026-09-28
  • #202 OpenSSF Scorecard: pin the remaining installs and re-check after 3.2.06 comments · 2026-09-28

Issue titles are quoted from GitHub. Open the link to check whether a problem still applies to the version you use.

05

Use cases

↳Coding with a local open model on your own GPU
↳Keeping source code off hosted AI services
06

Alternatives

⚡Top VS:
VSPI-Desktop★8.0k→VSzero★1.8k→VSopencode-telegram-bot★1.4k→VSdsh-Go★26.0k→
PI-Desktop logo
PI-Desktop★ 8.0k
Local-first AI coding agent desktop: Electron + Rust host core + pi Agent Harness + user-installable plugins
VSCompare→
CoreCoder logo
CoreCoder★ 1.9k
Minimal AI coding agent (~1,000 lines of Python) inspired by Claude Code. Works with any LLM. Think NanoGPT for coding agents. Formerly NanoCoder.
zero logo
zero★ 1.8k
The coding agent that answers to you, your model, your machine, your rules.
VSCompare→
opencode-telegram-bot logo
opencode-telegram-bot★ 1.4k
OpenCode mobile client via Telegram: run and monitor AI coding tasks from your phone while everything runs locally on your machine. OpenCode V2 support.
VSCompare→
dsh-Go logo
dsh-Go★ 26.0k
dsh-go — DeepSeek Harness 装进口袋:手机本机运行,也能遥控你的电脑。
VSCompare→
command-code logo
command-code★ 4.2k
Command Code AI — the best coding agent for open models.
VSCompare→
waku-agent logo
waku-agent★ 2.2k
Waku Waku! Waku Agent is a local-first AI agent harness you actually own, including loop, memory, eval, all in code built to stay legible as it grows.
VSCompare→
lilbee logo
lilbee★ 71
Chat with your documents offline using your own hardware. Augment any AI agent via MCP with hybrid RAG search over PDFs, code, and 150+ formats. Integrate with your favorite GUI via REST API.
VSCompare→
See all alternatives →

Comments

Log in to leave a comment

No comments yet. Be the first!

On this page
01Features02Why choose it03Trade-offs04Maintenance check05Use cases06Alternatives
inferstep /
ATLAS
↗

Local coding agent bringing frontier-style test-time reasoning and automated repair to compact open-source models

2.1k201AGPL-3.0
Last commit2d ago
StatusActive
CategoryCode Assistant
Contributors12
Latest releasev3.1.7
Links
Documentation↗Discussion↗Issues↗Releases↗
Cloud Infrastructure
Ad

Deploy on DigitalOcean

Host your MCP servers, Docker containers, and AI agents with $200 free credit over 60 days.

Claim $200 Free Credit
© 2026 AgentIndex.app|Built by a 10-year iOS Developer.
Best by use caseAboutContactAdvertisePrivacy PolicyTerms of UseQYSGitHubBuy me a coffee ☕

Browse by Category

MCP ServersCode AssistantWorkflow AutomationRAG / Knowledge BaseMulti-AgentBrowser AutomationLLM InfraDev ToolingObservability

Not affiliated with Anthropic, OpenAI or Microsoft.