EvoX Agent
An AI programming and productivity agent developed by EvoMap, available on Mac and Windows. Core features include swarm collaboration, self-evolving skills, and automatic multi-model switching. The agent can read codebases, write code, execute commands, and fix failing tests.
Timeline (last 90 days)
Swarm mode isolates each task into an independent agent, each with a clean context, and uses code rather than a model to merge results, thereby avoiding information loss during the summarization process.
EvoX Agent can connect to LM Studio, Ollama, and standalone devices such as DGX Spark via the AI services page to run local models such as Nemotron and Qwen.
EvoX Agent grants 1,500 free credits upon registration with no credit card binding required to try it out, and during installation it supports one-click import of settings, skills, and connectors from Codex and Claude Code.
EvoX offers three working modes: Chat, Cowork, and Code.
EvoX's self-evolving skills follow a five-step process: Observe, Discover, Reuse or Improve, Validate, and Consolidate, where a Gene is the smallest tested strategy and a Capsule is the packaged bundle of an entire task and its solution.
EvoMap claims that once EvoX learns common patterns in a codebase or conversation, token consumption can be reduced by approximately 31%.
Processing an Indonesian-language application task in swarm mode takes approximately 84,000 tokens and seven to eight minutes to complete, whereas turning off the swarm and executing sequentially takes approximately 2.5 times as long and consumes 129,000 tokens of context.
EvoX offers an optional global network that allows users to inherit high-scoring genes or capsules from other agents, publish their own results for others to reuse, and earn additional credits through contributions.
EvoX Agent's self-evolution mechanism consists of five steps: Observe, Discover, Reuse or Improve, Validate, and Solidify.
A Gene is the smallest validated strategy, a Capsule is a packaged solution for a complete task, and an Event is a non-editable audit log.
13 more timeline events
All Facts (20)
Swarm mode isolates each task into an independent agent, each with a clean context, and uses code rather than a model to merge results, thereby avoiding information loss during the summarization process.
50%UnverifiedEvoX Agent can connect to LM Studio, Ollama, and standalone devices such as DGX Spark via the AI services page to run local models such as Nemotron and Qwen.
50%UnverifiedEvoX Agent grants 1,500 free credits upon registration with no credit card binding required to try it out, and during installation it supports one-click import of settings, skills, and connectors from Codex and Claude Code.
50%UnverifiedEvoX offers three working modes: Chat, Cowork, and Code.
50%UnverifiedEvoX's self-evolving skills follow a five-step process: Observe, Discover, Reuse or Improve, Validate, and Consolidate, where a Gene is the smallest tested strategy and a Capsule is the packaged bundle of an entire task and its solution.
50%UnverifiedEvoMap claims that once EvoX learns common patterns in a codebase or conversation, token consumption can be reduced by approximately 31%.
50%UnverifiedProcessing an Indonesian-language application task in swarm mode takes approximately 84,000 tokens and seven to eight minutes to complete, whereas turning off the swarm and executing sequentially takes approximately 2.5 times as long and consumes 129,000 tokens of context.
50%UnverifiedEvoX offers an optional global network that allows users to inherit high-scoring genes or capsules from other agents, publish their own results for others to reuse, and earn additional credits through contributions.
50%UnverifiedEvoX Agent's self-evolution mechanism consists of five steps: Observe, Discover, Reuse or Improve, Validate, and Solidify.
50%UnverifiedA Gene is the smallest validated strategy, a Capsule is a packaged solution for a complete task, and an Event is a non-editable audit log.
50%UnverifiedEvoMap claims that after learning common patterns, approximately 31% of tokens can be saved.
50%UnverifiedIn EvoMap's benchmark test, on 563 questions using the same AI model, a single agent working independently achieved a 26% accuracy rate, while swarm collaboration mode achieved 71%.
50%UnverifiedIn the survival game Arena experiment, approximately 100 agents working in parallel completed the task in about 20 minutes, whereas sequential execution would have taken several hours.
50%UnverifiedIn the evolution phase of the survival game experiment, 64 species survived in the first generation; after the second generation discovered the strategy of 'staying still and initiating fewer attacks,' 99 species survived.
50%UnverifiedEvoX Agent can connect to channels such as Discord, LINE, Messenger, and Slack via Connections, and can also scan a QR code to link a mobile phone for triggering tasks.
50%UnverifiedIn a comparative test of an Indonesian-language flashcard application, swarm mode completed the task in approximately 7–8 minutes consuming about 84,000 tokens, while sequential execution took approximately 2.5 times longer and consumed about 129,000 tokens.
50%UnverifiedEvoX Agent is a coding and productivity agent launched by EvoMap, currently supporting Mac and Windows.
50%UnverifiedEvoX Agent has core capabilities including reading codebases, writing code, running commands, and fixing tests.
50%UnverifiedEvoX Agent supports importing settings from Codex and Claude Code, with skills and connectors migratable as well.
50%UnverifiedEvoX Agent grants 1,500 credits upon registration and can be tried without binding a card.
50%