One living source of truth for humans and agents.
Stable Baseline is the documentation and project workspace where every AI agent — Claude, GPT, Gemini, anything MCP — shares the same context and collaborates with your team against real timelines.
Tech Spec: Auth Service
DOC-115v51mo agoDocuments— Rich docs with inline diagrams, co-authored by agents.
Bring your own agent
Any agent,
one shared brain.
Claude Code, Cursor, Windsurf, Copilot, Warp, OpenAI Codex, ChatGPT, Gemini — whichever tool your team uses, it reads and writes against the same Stable Baseline workspace through MCP. No more copy-paste, stale context, or tool-lock-in.
Unified context
Every tool starts from the same canonical docs, diagrams, and plans.
Client-agnostic
Swap IDE or chatbot tomorrow; your knowledge graph stays intact.
Permission-aware
Fine-grained scopes per agent — read, write, diagram, or plan.
Documentation is the source
Living documents with
diagrams inline.
Documents aren't static markdown — they carry architecture diagrams, sequence flows, data tables, and media as first-class blocks, all edited by humans and agents in the same workspace.
A workspace as capable
as your team.
Documents, diagrams, and work — woven together by one knowledge graph so nothing falls out of sync.
Living documentation
CDMD — a Markdown superset with inline diagrams, data tables, and patches. Agents can edit via `findAndReplace` or line-based patches.
40+ diagram types
Mermaid, PlantUML, BPMN, D2, GraphViz, Excalidraw, ERD, sequence, and more.
Every change traced
Every agent and human edit is auditable. Roll back a section or the whole doc.
Humans + agents, roled
Scoped permissions per agent. Read-only reviewers, write-access authors, diagram-only workers.
Searchable knowledge
Full-text + semantic search across docs, diagrams, and plans — instantly.
MCP-native
60+ production tools. Works with Claude, GPT, Gemini — and every MCP client you'll use next.
Audit-ready documentation
Scan your docs against popular frameworks — OWASP, NIST, CIS — and log gaps as improvements automatically.
OWASP
app security
NIST SSDF
dev lifecycle
CIS v8
operational
Git-native workflow
Export, diff, and sync with your repo. Docs live where your code does.
Synced to main · 2 min ago
Plans, phases, tasks
Project work
on a shared timeline.
Every plan has phases, every phase has tasks — each one assignable to a teammate or an agent. Zoom across Day, Week, Month, Quarter, or Year, and park work in the unscheduled lane until it's ready.
WBS hierarchy
Auto-numbered work-breakdown codes for every phase and task.
Drag to reschedule
Resize or move bars; start and end dates update on drop.
Unscheduled lane
A parking section for tasks that aren't dated yet.
Q2 Auth rollout
Captured, triaged, shipped
Never lose an idea,
never miss a risk.
Every gap an agent notices and every idea a teammate raises lands in one backlog. Statuses walk from captured to shipped, with compliance evidence linked to source docs automatically.
Risk types tracked
7
bug · risk · gap · feature
Status workflow
10
captured → shipped
Priority tiers
4
low · med · high · critical
Evidence linked
auto
docs · diagrams · plans
ComplianceEnterprise
Catch gaps early,
before auditors do.
Scan the documentation you already have against popular security and privacy frameworks. Stable Baseline highlights where evidence is missing — so you go into formal audits with nothing to scramble for.
A readiness pass that cross-references your workspace docs with public framework catalogues — and logs what's missing as improvements.
OWASP ASVS 5.0
Application securityNIST SSDF 1.1
Secure dev lifecycleCIS Controls v8
Operational securityISO 27001 · SOC 2 · TISAX
Framework catalogues growingCompany Brain · Knowledge GraphEnterprise
One brain. Every agent.
Every document, plan, and improvement your team writes is fed — automatically — into a living company brain. Claude, Cursor, ChatGPT, Gemini, internal agents: they all read the same brain over MCP. No skill files to maintain. No per-tool context drift. The brain self-learns as your workspace grows.
→What is this project about? Summarise the main themes.
No skill files. Ever.
Stop maintaining one context file per agent. Every connected agent — public (ChatGPT, Claude, Gemini) or internal — reads the same brain via MCP. Update your docs once; every agent sees it instantly.
Fully automated · self-learns
The brain rebuilds itself on every save. New documents, plans, improvements, diagrams — all parsed, linked, clustered, summarised in the background. Zero curation. Zero ETL.
Right answers, faster — and honest when there isn't one
Graph traversal returns a tight, grounded neighbourhood — entities, relations, and summaries — so agents reach the right answer in fewer hops. Just as importantly, when the brain has nothing on a question, agents say so instead of hallucinating. Less guessing, more grounding.
One brain — or one per department
Run a single workspace-wide brain, or segment by function: HR, Finance, Operations, Engineering, Sales — each in its own workspace or project. Departments stay isolated by default and shareable on demand.
RBAC end-to-end
Every read goes through the same scope walk Stable Baseline uses for documents — org → workspace → project → folder → document. Agents inherit the caller's permissions exactly. No bypass paths.
Interactive WebGL Explorer
A WebGL canvas that renders your whole brain. Pan, zoom, click any node to focus, follow a relation, or pivot to its neighbourhood. Useful for onboarding, audits, and just seeing what the company knows.
Available on Enterprise. Strictly opt-in: org → workspace → project, with a four-gate scope check on every code path.
Built-in agent
Meet the Stable Baseline Agent.
No setup, no API key, no config file. Open any doc, project, or workspace and the agent is right there — creating docs and diagrams, updating plans, phases, tasks, and improvements, answering questions, and navigating on your behalf. Drop in a standup transcript and it reshapes the plan, assigns tasks, and drafts comments — context handled for you.
Understands your scope
Opening a document narrows the agent to that doc. On a project view, the scope expands. No IDs, no copy-paste — context is automatic.
Drives your plans & tasks
Paste a meeting transcript and watch it land: phases re-scheduled, tasks assigned to the right people, comments drafted, improvements logged — all threaded against the project it belongs to.
Frontier models, handled for you
We route through zero-data-retention providers and pick frontier-class models so you always get the best answer. You don't bring a key; you don't pick a model; you just describe what you want.
Fair, transparent billing
Chatting with the agent is free. Only the heavier calls — web fetches, file uploads + parsing, deep research — deduct credits from your balance on the fly, shown live in the chat.
Solutions for every team
Built for the way your team works.
One living knowledge graph, tailored views for every role — project managers, product teams, support, and business users.
Project managers
Plans, people, and agents on one canvas.
Run every initiative off a single plan where humans and agents share phases, tasks, and status. Close the gap between what's planned, what's shipping, and what the docs say.
- See status live — no weekly roll-ups or status meetings
- Assign tasks to teammates or agents from the same view
- Every task links to the doc, diagram, or improvement it moved
Q2 Platform Release
PLN-0042 · 4 phases
Discovery · 3 tasks
100%Design · 5 tasks
72%Build · 8 tasks
38%Launch · 2 tasks
0%3
Agents
5
Humans
12 tasks
This week
Product & engineering
Specs that evolve with the product.
Architecture diagrams, sequence flows, and API specs live next to the decisions that produced them — and your agent keeps them accurate as the code changes.
- Inline diagrams, not flat screenshots — source is always editable
- Decisions captured with evidence, linked to the code that shipped
- Version history across every edit, human or agent
API: Auth Provider
Decision log
- Use PKCE for mobile clientsClaude · v12
- Drop session fixation fallbackSarah · v11
- Switch to Redis for noncesGPT · v10
Support teams
Lightning-fast agent-assisted support.
Your support agent reads from the same knowledge graph your product team writes to. Every ticket becomes a triaged improvement the moment it lands.
- Front-line agents answer from canonical docs, never out of date
- Tickets auto-convert to improvements with severity and evidence
- Customer context carries across handoffs — no more re-explaining
Ticket #4821
3m agoAuto-triaged
IMP-7291
Migrate legacy auth tenants to PKCE
Evidence
- DOC-115 · §3.2
- Ticket #4821
- OAuth2 RFC 6819
Business users
Your context, always with you.
Personal docs, meeting notes, research, and decisions all flow into one space an agent can reason about. No more fragmenting your thinking across five tools.
- Your AI knows your docs, not just the last message you typed
- Private and team spaces — share by intent, not by tool
- Work with any model; your knowledge stays yours
Vineet's workspace
- Q2 strategy notes
- Customer interviews
- Competitor analysis
- Founder 1:1 log
- Weekly reflections
- Goals · 2026
Reading your context
live“Based on your Q2 notes and last week's customer interviews, three retention drivers repeat across 6 segments — worth expanding in your board deck.”
Q2 strategy notes
“retention is our lead metric”
Customer interviews · Acme
“they want recurring context”
Goals · 2026
“reach 45% MoM retention”
Your data — private, yours, portable.
Diagrams, not drawings
40+ diagram types,
text-first and versioned.
Every diagram lives as source. Review diffs, roll back versions, generate new ones through your agent. Export to SVG, PNG, or PDF — the canonical form is always plain text.
BPMN · lanes
Business process
ERD
Data model
Sequence
Messaging flow
Flowchart
Decision tree
Charts · Vega
Data-bound
GraphViz
Directed graph
Live examples · category filters · free-tier toggle
Mermaid · PlantUML · BPMN 2.0 · D2 · GraphViz · TikZ · Excalidraw · Vega · Sankey · ERD · WBS · Gantt · Mind maps · and many more
Where to find us
Stable Baseline, wherever you work.
Connect through any MCP client, or open one of our published apps to bring Stable Baseline directly into the tool your team already lives in.
Shipping next
Claude
Apps directory
Coming sooniPhone & iPad
App Store
Coming soonAndroid
Google Play
Coming soonWindows
Microsoft Store
Coming soon