AI industrial development environment · for chip and embedded R&D
Connect AI
to chip and embedded
toolchains.
Describe what you need in natural language in AI clients such as Codex and Claude Code. Tai-IoT provides one consistent set of engineering tools, runs each platform's build and test tools, and returns reviewable changes and verification records.
- Runs locally; project code never passes through Tai-IoT servers
- Built on open standards; works with your existing AI clients
- Every change comes with build, test and diff records
Engineer's request · original in Chinese“……先读取工程上下文和 C 源码，把采样周期从 1000 毫秒改为 500 毫秒，报警阈值改为 28℃。实际编译，执行模拟测试，展示 Git 差异。……”“… First read the project context and the C source, change the sampling period from 1000 ms to 500 ms and the alarm threshold to 28 °C. Build it for real, run the simulated test and show the Git diff. …”
get_project_profilehost-simulator · 1000 ms / 60 °Cread_project_filefirmware/telemetry.cread_project_fileproject.jsonset_sampling(500, 28)Changed; rebuild neededbuild_project/usr/bin/cc build passedrun_simulation7/7 checks · 3 alarmsget_git_diffM project.json
- "sample_period_ms": 1000,
- "alarm_threshold_c": 60.0
+ "sample_period_ms": 500,
+ "alarm_threshold_c": 28.07 / 7 checks passed
Sampling period, clock wraparound, alarm boundaries
10 samples: 0–4500 ms, 500 ms apart
3 alarm samples: 28.0 / 28.5 / 29.0 °C
Clock: simulated / Temperature: synthetic data7 tool calls · real host build · 7 simulated checks passed · 1 min 30 s
A real run on the example project: the model chose the tools itself, with no human intervention. Full record →
How it works
One request, reviewable results
Describe the task in the AI client you already use. Tai-IoT carries out the engineering work on your computer and hands the results back for you to confirm.
- 01 / Describe the task
Say what needs to be done
In an AI client such as Codex or Claude Code, describe the change and how it should be verified.
- 02 / Tai-IoT
Engineering work on your computer
Reads the project, makes the change, builds and tests it, all on your own computer.
- 03 / Review
Deliver after you confirm
Review the diff and the build and test records. Any step that did not pass is explained, and an engineer confirms before delivery.
Platform adaptation
For Chinese and international chip platforms
115chip vendors (incl. peripheral devices)
toolchain information compiled
Development kits, compilers, download and debug tools and engineering software interfaces have all been reviewed. Support packs are delivered per project: adapted to the target model, development tools and versions, and delivered with build and test records. The development board example in the trial toolkit is cross-compiled with the vendors' official SDKs, and its build records are published in the release notes.
PLC projects: Each PLC brand, and often each series within a brand, uses its own programming software, and the project formats and download protocols belong to the vendor. Tai-IoT works through the official interface provided by the vendor's engineering software and never bypasses it, so PLC work is evaluated project by project: it applies to platforms whose vendor software offers an open engineering interface, uses the customer's own licensed copy of the software, and an engineer confirms before anything is downloaded to a device.
Get started
Try the Tai-IoT toolkit for free
Connect your existing AI client on your own computer and, with a single request, read the project, change parameters, build for real, run tests and review the Git changes. After you apply online, the download links are sent to your email automatically.
Free trial
The trial toolkit is free for 90 days. Your AI client and model service use your own accounts, and model costs are your own.
Ready-to-run example projects
The host example is a temperature acquisition C program, so you can try the full change, build, test and review flow without a development board; the development board example is the same task as projects for three boards, cross-compiled with the vendors' official SDKs, with flashing and serial read-back tools. The tools work only on the included examples; to connect your own project, contact us.
Linux · VS Code · popular AI clients
The host example requires Python 3.11+, Git and a C compiler and was tested on Linux with Codex CLI; Claude Code and other clients that support MCP connect the same way. The development board example includes a Windows and VS Code guide (in Chinese), so you can use your own model service in VS Code. The toolkit includes English guides, and we are glad to help in English by email.
- 01
Apply
Enter your company, system and AI client online
- 02
Receive
Accept the terms; the download links arrive by email
- 03
Install
Verify, unpack, install dependencies, check the setup
- 04
Automatic demo
Confirm builds and tests run, without AI
- 05
AI task
Connect your client and give the task once
- 06
Feedback
Reply by email with blockers and adaptation needs
Success criteria: for the host example, the automatic demo and the AI task should both give 7 simulated checks passed, 10 samples and 3 alarm samples, and the AI task's Git diff shows the sampling period 1000 → 500 ms and the alarm threshold 60 → 28 °C, with simulated input and no hardware. For the development board example, the build result should show the compiler received the values for 500 ms and 28 °C.
Setup and walkthrough
Start from a local project
See the supported clients and setup options, then walk through installation, running a task and reviewing the results.
Installation and client setup
Choose the script, Codex CLI or Claude Code path, then view and copy the steps for your computer.
Open the installation guide →ENGINEERING WALKTHROUGHWorkflow walkthrough
From preparing the environment and connecting a client to describing a task and reviewing the results, step by step.
See the workflow →Use cases
Built around real engineering tasks
Typical work in industrial communication and embedded R&D. The approach depends on the specific model, development tools and project conditions.
Industrial commsDevice protocol integration and error handling
“Connect the temperature meter according to this register map, handle timeouts and reconnection, and add tests.”
- Inputs
- An existing MCU or gateway project, the device protocol, the register map and the development tools.
- Process
- Check addresses, data types and byte order, adjust the data mapping and communication logic, and test normal responses, timeouts and recovery paths.
- Delivers
- Component changes, build records, test inputs and results, plus the items that need on-device verification.
Fewer omissions and less back-and-forth when organizing protocols, writing repetitive code and checking edge cases.
Embedded devMCU feature iteration and regression testing
“Add fault-code handling to the acquisition module, keep the existing interfaces, and verify both normal and abnormal input.”
- Inputs
- The chip or board, SDK, compiler, project version and existing tests.
- Process
- Read the interfaces and configuration, limit the scope of the change, run the matching build, and verify the new logic and existing features.
- Delivers
- Code diff, build and regression results, tool version records, and the conditions and status of on-device tests.
Less time spent finding project information, organizing repeated tests and writing change notes.
TeamworkEnvironment recovery and project handover
“Restore the project environment on this computer and confirm that the given version of the project builds.”
- Inputs
- The repository, environment description, dependency sources and required third-party software licences.
- Process
- Diagnose what is missing, prepare the tool versions the project requires, and run the build and existing checks.
- Delivers
- An environment list, dependency records, reproducible steps and build results.
Less repeated troubleshooting when setting up new computers, handing over work or maintaining older projects.
MaintenanceBuild failure analysis and verified fixes
“Analyze this build failure, find the cause, propose a change and verify it again.”
- Inputs
- Failure logs, build configuration, tool versions and access to the project source.
- Process
- Separate dependency, configuration and source problems, propose a targeted change, rebuild and run the relevant checks.
- Delivers
- The evidence behind the diagnosis, the diff, the re-verification results and any open issues.
Less log collation and trial and error, with conclusions tied to real run results.
Enterprise R&D
Fits your existing R&D process
Deployment, versioning, review and maintenance options built around your own projects and toolchains.
Deployment and model connection
The engineering tools run in the environment your project requires, and the model connection follows your company's policy. When cloud models are used, the scope of project context sent is clearly defined.
Projects and tool versions
Environments are organized by repository version, SDK, compiler and target hardware, giving collaboration, handover and issue reproduction a shared basis.
Authorization and change review
Access and operations are limited to the task. Code changes, hardware connections and device operations follow your team's authorization, review and release process.
Task records and maintenance
Task status, build output, test conditions and error logs make each run traceable and keep the evidence needed to reproduce and maintain it.
About Tai-IoT
A development environment for industrial R&D
Tai-IoT® is the product brand of Hainan Taiyue IoT Technology Co., Ltd. (泰岳物联). It serves enterprise R&D teams and industrial IoT projects, provides unified MCP access to the development toolchains of Chinese and international chips, and builds an AI industrial development environment around project environments and verification workflows. Hardware adaptation follows each customer's project needs, and every support pack states the applicable models, tool versions and verification records.
Contact and trial questions
Start with your engineering needs
Tell us about your current projects, target platforms and the R&D tasks you want to improve. We will confirm the integration approach and verification conditions for your version scope.