Yunce AI
EN中
Contact
CompanyMethodTrustInsightsContactTurn intent into action
AGENTIC AI COMPANY · SHENZHEN

Build the agent
execution system.

Yunce AI builds applied AI and agentic systems — from goal understanding and tool use to workflow execution and verifiable, evidence-backed outcomes.

Explore the companyView execution model
Intent → ActionWhat we build
Bounded · VerifiableHow it runs
Human-in-the-loopAlways
EXECUTION TRACErun · a8f2c1
  • Model the intentparsed
  • Draw the boundarybounded
  • Orchestrate actionrunning
  • Verify the outcomeverified
09:24:02intent.parsed — "reconcile Q2 invoices"
09:24:03scope.bounded — 3 tools · 1 approval
09:24:18evidence.attached — outcome verifiable
Agentic AI
Company focus
Execution
Core theme
Evidence
Design principle
Shenzhen
Based in
COMPANY THESIS

A company building for the agentic application layer.

Yunce AI focuses on applied AI, agent orchestration, workflow automation, and trustworthy execution patterns that keep boundaries, approvals, and evidence visible. We don't claim AI replaces people — we make useful autonomy controllable, observable, and recoverable.

// Turn intent into action.

CAPABILITY PILLARS

Agentic applications need execution design, not just model access.

Four capability directions, designed around real workflows — from how an agent understands a goal to how its results are proven trustworthy.

01

Agent Experience

Design the surfaces, context, and human handoffs that make AI agents usable in real work.

02

Tool Orchestration

Connect models with tools, data, APIs, and business actions through controlled execution paths.

03

Workflow Automation

Move repetitive, cross-system, standardizable work forward while preserving human judgment.

04

Verifiable Execution

Preserve boundaries, approvals, logs, and evidence so AI-assisted work stays trustworthy.

WORKING METHOD

The execution model

See the full method
01

Model the intent

Clarify the actual work, success criteria, and the output that matters.

02

Draw the boundary

Map data, permissions, risk, human approvals, and what should not be automated.

03

Orchestrate action

Combine models, tools, workflows, and interfaces so agents can move within bounds.

04

Verify the outcome

Use logs, evidence, replay, and human review to decide whether the result is trustworthy.

TRUST SURFACES

Future-facing systems still need control surfaces.

The point is not to make AI look autonomous. The point is to make useful autonomy observable, bounded, and recoverable — with evidence attached to every outcome.

Bounded tool permissionsscope.bounded
Human approval for sensitive actionsapproval.required
Observable execution statestate.observable
Recoverable failure pathsfailure.recoverable
Evidence attached to outcomeevidence.attached
RESEARCH NOTES

Insights & Research

Our thinking on AI agents, execution systems, workflow design, and real-world implementation — methods and judgment, not trend chasing.

PLAYBOOK · 6 MIN

Why teams need an AI agent execution layer

The shift from chat assistants to trustworthy execution systems depends on boundaries, logs, approvals, and proof of completion.

AI AgentExecution LayerWorkflow
Read all insights
EVIDENCE CHAIN4 / 4
  • Goaldefined
  • Approvalgranted
  • Actionlogged
  • Proofattached
FOUNDER · OPEN SOURCE IMPACT

Built by an open-source engineer whose work is validated by the community.

Yunce AI is led by 程序员阿江-Relakkes, an engineer with 10 years of software development experience, large-company experience, and a visible track record in open-source infrastructure, data crawling, AI agents, developer tooling, and applied AI education.

His open-source and research-grade engineering projects include MediaCrawler, a 50K+ star multi-platform crawler project, and cc-haha, a 12K+ star developer-tooling project. This is the kind of community-validated engineering signal investors and venture teams can inspect directly: real users, real forks, real technical discussion, and shipped systems rather than resume-only credibility.

GitHub profileMediaCrawlercc-haha
程序员阿江-Relakkes
程序员阿江-Relakkes

Founder · Open-source engineer · AI builder

50K+StarsMediaCrawlerMulti-platform crawler project used by a large developer community.10K+ forks12K+Starscc-hahaDeveloper-tooling and AI coding workflow project with strong community attention.8K+ forks
Development
10 years of software development experience
Leadership
Former Tech Lead and Technical Expert roles
Education
Self-media AI education across models, agents, and engineering practice
Implementation
Hands-on AI implementation across workflows, tools, and execution systems
GET STARTED

Planning an AI agent or applied AI initiative?

Start with one concrete workflow. We can help reason through the objective, boundary, execution path, and verification model.

Contact Yunce AIView execution model