๐จ New Feature alert !!!
Introducing Code Review in Bahulam โ configurable AI reviews for your GitHub repositories.
โก๏ธ Custom instructions and coding standards for each repository, editable anytime.
โก๏ธ Specialist reviewers: security, bugs, PR summaries, coding standards, function documentation, and consistency with similar features.
โก๏ธ Related-code checks across changed functions, callers, dependencies, tests or any other changes in the PR.
โก๏ธ Contextual or deep review, with your architecture and standards files as reference.
โก๏ธ Configurable documentation coverage targets for new or modified functions.
https://t.co/hkPRoEw3it
#Bahulam #CodeReview #DevTools
๐๏ธ๐๏ธ Voice is now available in Bahulam.
Tap the mic in Chat, mobile, or a supported workspace, speak naturally, then review and edit the transcript before you send it.
Same agent. Same tools. A more natural way to get work moving.
Try it: https://t.co/mqw3liuL3y
#Bahulam #VoiceAI #AIAgents
โณ๏ธ Fun but useful Plugin
Now Bahulam can write a complete Research Document or Patent Application from the knowledge.
1. cd into the folder with the documents
2. `Bahulam plugin document-writer .` --> Opens the document-writer workspace
3. ask Bahulam to write a research paper and iterate till you are ready to publish.
visit ๐ https://t.co/49TQttevXe
What differentiate Bahulam Plugins to Claude Plugins (MCP)
1. Mini-Products vs. Raw Pipes ๐งฉ
Claude MCP servers are the โpipesโ that connect data to a model. Bahulam Plugins are Mini-Products. We bundle the Agent (the expertise), the Tool (the pipe), and a Custom UI into a single manifest. You don't just get a connection; you get a specialized cockpit for the task.
2. Consolidated vs. Tool Sprawl ๐ ๏ธ
Instead of 10 fragmented tools, Bahulam Plugins can consolidate all the different tools into one plugin. What you get is one Plugin that is working on Outcome instead of multiple tools making agent patch
Build the Outcomes not Applications that you can forget
#AI #MCP #BahulamCode #BuildInPublic #AgenticWorkflows #ClaudeAI #OwntheOutcome
Don't build applications.
Build Outcomes
Bahulam Plugins will help build plugins that get the outcomes with AI rather than just running behind Applications.
now you can use the plugins to create ThreeJs animations.
One bundle. Tools + Sub Agents + Workspace Views for one use case.
Install, Publish, share with the community Build Outcomes, not loose pieces
https://t.co/VPYZaLeMYx
#pidev #Hermes #DevTools #AICoding #opensource
๐ Introducing DeepSeek-V4.1-Flash: smarter, faster, more efficient.
๐น Introducing the smallest model in our new architecture family, with native visual understanding.
๐น Designed for greater capability, faster inference, higher throughput, and scaling to larger models.
1/6
Bahulam/Code v0.1.19 introduces a cleaner sub-agent workflow for real workspace tasks.
The primary agent stays in control, while specialists can explore, plan, verify, debug, or handle domain-specific work when needed.
Sub-agents are:
- configured in YAML
- triggered by task, tool, or command context
- easy to extend with your own specialist prompts and tools
This is the foundation for open, plugin-driven workspaces where teams can bring their own agents without replacing the platform agent.
#Bahulam #AIagents #DevTools #OpenSource #Harness
Why settle for a "headless tool" when you can deploy a "mini-product"? ๐งฉ
At https://t.co/hkPRoEw3it, we are moving past the era of generic AI plugins. Most AI tools today fail because they are just raw APIsโthey expect YOU to be the expert prompter to get results.
The architecture shifts the burden from the user to the system using a Mini-Product Strategy.
Every Bahulam Plugin is a coherent bundle consisting of the "Trinity of Capability":
1๏ธโฃ The Agent: A specialized expert persona. You don't prompt it; you collaborate with it.
2๏ธโฃ The Tool: High-performance MCP-powered functions (code execution, data rendering, API logic) that run in a secure, sandboxed environment.
3๏ธโฃ The UI: Custom workspace panels that provide a visual cockpit. You don't just "chat" with a databaseโyou interact with a dashboard.
The Benefit? Portable Expertise. ๐
Because we use a unified manifest, your plugins aren't locked to one screen. Start a complex data animation on your CLI, monitor the render via Web, and get the final result as a push notification on Mobile.
By bundling the Agent + Tool + UI, we aren't just giving you a "skill"โwe are giving you "Expertise-in-a-Box" that works wherever you do.
Stop chatting with APIs. Start deploying capabilities. ๐
#AI #Agents #MCP #Developers #BahulamCode #OpenSource
Building with https://t.co/2D8r7ev2JI? You can now wrap it as a Bahulam plugin in an afternoon:
The plugins let you build outcomes, not applications that lead you nowhere.
1๏ธโฃ Expose https://t.co/2D8r7ev2JI tools via RPC/SDK โ MCP server
2๏ธโฃ Create a Bahulam plugin: tools + sub-agent + workspace view
3๏ธโฃ Wire Shared Blackboard for state
4๏ธโฃ Publish and share
npm i -g @bahulam/code โ https://t.co/hkPRoEw3it
Show us what you build. ๐
#BuildInPublic #OpenSource #AIagents
@adhamuxi Try Bahulam/code .
It uses multiple models for different tasks preserving context and give more usage
You can also choose the model depending of the session so that you can switch between exploration read and write tasks
https://t.co/1JvCElFlWh
Why pair Bahulam/Code
https://t.co/VPYZaLeMYx
with https://t.co/2D8r7ev2JI?
๐ง https://t.co/2D8r7ev2JI: 15+ providers, hundreds of models, mid-session switching, tree sessions
๐ก๏ธ https://t.co/wN38cV35p2: HITL approvals, local checkpoints with undo, session-scoped permissions
You get https://t.co/2D8r7ev2JI's model flexibility inside @BahulamAI's reliability + safety layer.
Power without losing control. โ๏ธ
try ==> npm i -g @bahulam/code
every new user gets 250 credits to try
#AIcoding #BuildInPublic
@PeilvDog_CN Yes. Valid suggestions. Right now plugins have tools with schemas pre defined so that agent call can be deterministic.
Persisted run ID is very good suggestion. Will try to update in the next version
How a https://t.co/wN38cV35p2 + https://t.co/2D8r7ev2JI plugin works:
1๏ธโฃ https://t.co/2D8r7ev2JI exposes its tools via RPC/SDK โ MCP server
2๏ธโฃ https://t.co/Xw28tCqeZy plugin registers those as callable tools
3๏ธโฃ Sub-agent (system prompt + tool whitelist) delegates coding tasks to pi
4๏ธโฃ Results land in Shared Blackboard SQLite
5๏ธโฃ Workspace view renders live via SSE
TypeScript end-to-end. No glue code. ๐งต๐
browse availble plugins.
https://t.co/VPYZaLeMYx
#DevTools #TypeScript #AIagents
Create your own Manim Animation in 3 steps
๐ install BAHULAM --> npm i -g @bahulam/code
๐ install Manim Animation Plugin --> bahulam install bahulam:manim-studio
๐ open the studio and ask the agent --> bahulam plugin manim-studio .
You can ask the agent by describing what animation you want.
Visit https://t.co/kgUzNrIFXK for more details
#AiAnimation #manim #pi
What happens when two terminal-native, TypeScript-first agent platforms connect?
๐ค Bahulam Plugins ร https://t.co/2D8r7ev2JI
https://t.co/2D8r7ev2JI has several useful plugins that Bahulam wraps it as a sub-agent inside a plugin.
Results flow back through the Shared Blackboard with live workspace views.
One plugin. Two harnesses. Zero glue. ๐
install with npm i -g @bahulam/code
install any plugin using bahulam install pi:<plug-in>
https://t.co/hkPRoEw3it
#AIcoding #MCP #BuildInPublic
Feature Update - v0.1.16
Now Bahulam/Code can delegate multiple sub agents at the same time and explore, plan, review the codebases.
Try
npm i -g @bahulam/code
Bahulam Plugins as Live
One bundle. Tools + Subagents + Workspace Views for one use case.
โ Runs client side
โ Shared Black board SQL Lite state with Live SSE events
โ MCP integration (stdio, SSE, WebSocket, HTTP)
Install, Publish, share with the community
Build work flows, not loose pieces
#pidev #DevTools #AICoding
https://t.co/MMph9DYdQf