Chart types
line, bar, pie, radar
Developer Preview
Pipintama is becoming a hosted tools platform for autonomous agents.
Boards is live today for mindmaps, flowcharts, kanban boards, and architecture views. Charts is the next tool on the same platform for line, bar, pie, and radar visualizations. Agents do not need your repo. They call your server, create the artifact, and return a hosted link.
Next tool
Pipintama Charts
Charts extends the same workspace, API key, MCP, and PNG export model into quantitative outputs. The initial plan uses Chart.js for a fast MVP with line, bar, pie, and radar chart types.
- Agent asks for a chart instead of a board when the task is about values, trends, or distribution
- The platform returns a hosted chart URL plus PNG when needed
- The auth, usage, and client attribution model stays shared with Boards
Renderer
Chart.js for the MVP export and hosted view
Primary output
viewer_url + png_url for chat channels
MCP path
new create_chart / get_chart / export_chart_png surface
What exists now
- Hosted viewer at `boards.pipintama.com`
- API at `api.pipintama.com`
- Platform at `pipintama.com/platform` with auth, workspaces, and API keys
- Visibility modes: private, shared, public
- Viewer modes: mindmap, flowchart, kanban, architecture
How agents use it
- Pick the correct board mode
- Call the Boards API or MCP layer
- Create the board in a workspace
- Return a hosted viewer URL to the user
- Return a PNG export too when the channel benefits from an image
Authentication now
- The hosted MCP requires a Pipintama API key
- Clients authenticate with `Authorization: Bearer <key>` or `x-api-key`
- Usage is attributed to the client id and name, not only the IP address
- OAuth is reserved for a later multi-user phase across future Pipintama tools
Board Modes
Each mode is optimized for a different kind of request. The screenshots below come from the current live product in light and dark themes.
Mindmap
Use for concept expansion, brainstorming, topic maps, and idea clustering.
- One clear root concept
- Short branch labels instead of paragraphs
- Best for exploration, not operational processes
Light mode
Dark mode
Flowchart
Use for processes, approvals, branching logic, and yes/no style decisions.
- Prefer start, process, decision, and end semantics
- Use branching only when there is a real condition
- Keep each step action-oriented and readable
Light mode
Dark mode
Kanban
Use for tasks grouped by stage such as backlog, doing, review, and done.
- Each card should be a concrete task
- Lanes should represent status or workflow stage
- Best for execution and coordination, not system design
Light mode
Dark mode
Architecture
Use for services, gateways, stores, integrations, and system dependencies.
- Prefer short component names
- Show dependencies clearly
- Use for systems and infrastructure, not human workflows
Light mode
Dark mode
MCP Surface
The intended model is not a public repo workflow. Agents connect to your server and call stable tools.
Primary tools
create_board
Generate a hosted board from user text and return a viewer URL.
get_board
Fetch metadata, scene JSON, and the hosted viewer URL for an existing board.
share_board
Create a shared tokenized link for board access.
set_board_visibility
Switch a board between private, shared, and public access.
update_board
Refine an existing board without creating a new board id. Useful for agent-driven iteration.
export_board_png
Return a PNG export URL when the channel needs an image as well as a hosted link.
Access rules
- `shared` should remain the default for agent-created boards
- Use `public` only when the user explicitly wants an open link
- Use `private` only when authenticated workflows exist
- Require an MCP API key for remote client access
- Return the hosted viewer URL before offering raw JSON
Live endpoint
Agents should connect to https://api.pipintama.com/mcp, infer the mode, create a concise title, preserve the user’s source text, and return a hosted board link that can be opened immediately.
MCP endpoint
https://api.pipintama.com/mcp
Streamable HTTP endpoint for remote agents. Requires a Pipintama API key.
MCP health
https://api.pipintama.com/mcp-health
Simple public check for service availability.
PNG export example
https://api.pipintama.com/mcp-exports/cmndns0hn0001o401md5okzju.png?theme=light
Direct image export for channels like Telegram or WhatsApp.
Agent skill
/docs/pipintama-boards-skill.md
Markdown instructions agents can read before using the tools.
GitHub docs
https://github.com/davidk2yoyo/pipintama_agents
Public MCP and skill reference repo for humans and agent builders.
Charts spec
/docs/pipintama-charts-spec.md
Initial product and MCP contract for Pipintama Charts.
Privacy
- Pipintama logs client id, route, IP, user-agent, status, and timing for reliability and abuse protection
- Shared and public boards should be treated as link-accessible content
- Private boards are intended for authenticated client workflows
Safety
- MCP tools expose safety annotations such as read-only and idempotent hints
- Agents should prefer hosted boards over ASCII fallbacks when access is available
- Public links should only be created when the user explicitly asks for them
Support
- For access or support, start at pipintama.com/contact
- The current access model is platform API keys issued by Pipintama
- OAuth is planned for future user-scoped permissions across multiple tools
What comes next
The hosted MCP layer is now live for Boards, and Charts is specified next on the same auth and workspace model so Pipintama can become a real platform of hosted tools for autonomous agents.