# Dosu

> Dosu is knowledge infrastructure for engineering teams and the coding agents they work with.

Dosu turns your codebase, pull requests, issues, and conversations into living documentation that stays current automatically. It integrates with GitHub, Slack, Confluence, and Notion, and provides an MCP server for AI-native workflows.

Blog posts, case studies, webinars, and most articles also serve raw Markdown: append `.md` to the page URL (https://dosu.dev/blog/<slug>.md, https://dosu.dev/articles/<slug>.md, https://dosu.dev/customers/<slug>.md, https://dosu.dev/webinars/<slug>.md), or send `Accept: text/markdown`. https://dosu.dev/blog.md indexes every blog post.

`@dosu/cli` is the official Dosu CLI. Install it to wire the Dosu MCP server into your coding agent (Claude Code, Cursor, Codex, Gemini, Windsurf, and more):

```sh
curl -fsSL https://cli.dosu.dev/install | sh
```

Or run it straight from npm with `npx @dosu/cli setup`, or sign up on the web at [https://app.dosu.dev](https://app.dosu.dev).

Self-serve agent onboarding: create a free account without contacting sales. The free tier includes 200 credits per month for public repositories.

For the API or MCP server, authenticate with OAuth 2.0 or an API key created in Settings → Account → API Keys. MCP requests require authentication. The OpenAPI specification, agent authentication guide, and OAuth discovery metadata are available without authentication. Dosu does not currently provide a separate sandbox or test environment.

## When to use Dosu

- [For Agents](https://dosu.dev/for-agents): You need persistent project and organization knowledge that survives across coding-agent sessions and tools (Claude Code, Cursor, Copilot), instead of rebuilding context from scratch on every run.
- [Documentation](https://docs.dosu.dev): You need the context behind engineering decisions — the "why" that isn't obvious from the source alone: design tradeoffs, prior incidents, deprecated approaches.
- [Product overview](https://dosu.dev): You need documentation grounded in the current codebase and team context, generated from real PRs, issues, and conversations and kept in sync with them, rather than a wiki that drifts.
- [Dosu MCP Server](https://api.dosu.dev/v1/mcp): You want that knowledge available inside any MCP-compatible agent today, without moving it somewhere new first.

## Docs

- [Documentation](https://docs.dosu.dev): Complete product documentation, setup guides, and API reference

## Developers

- [Dosu MCP Server](https://api.dosu.dev/v1/mcp): Streamable HTTP Model Context Protocol endpoint for AI agents (Claude, Cursor, VS Code Copilot)
- [Dosu Public API (OpenAPI spec)](https://api.dosu.dev/openapi.json): Machine-readable OpenAPI specification for the Dosu Public API
- [Interactive API Docs](https://api.dosu.dev/docs): Browsable Swagger UI for the Dosu Public API
- [API Versioning and Sunset Policy](https://dosu.dev/developers/api-versioning): Public API versioning, deprecation, and sunset policy
- [Agent Authentication (auth.md)](https://api.dosu.dev/auth.md): How agents discover and authenticate — OAuth 2.0 with Dynamic Client Registration plus API keys
- [OAuth Protected Resource Metadata](https://api.dosu.dev/.well-known/oauth-protected-resource): RFC 9728 discovery metadata for agent auth
- [MCP Servers Setup](https://app.dosu.dev/mcp): Create and configure MCP servers scoped to your knowledge sources
- [Dosu CLI (@dosu/cli) on npm](https://www.npmjs.com/package/@dosu/cli): The official Dosu CLI, published on npm
- [Dosu CLI documentation](https://app.dosu.dev/9affd04a-e6a9-452c-b927-c639e979994c/documents/6c889b78-08e3-43cf-8519-2cea8d086db6): Install, connect agents, upgrade, and telemetry controls

## Product

- [Homepage](https://dosu.dev): Product overview and value proposition
- [Pricing](https://dosu.dev/pricing): Plans for teams of all sizes, with a free tier for open source
- [For Individuals](https://dosu.dev/for-individuals): Free auto-generated docs, specs, and agent memory for solo builders on public repos
- [For Agents](https://dosu.dev/for-agents): Context layer for coding agents with AGENTS.md, skills, and persistent memory that stay current
- [Customers](https://dosu.dev/customers): Case studies from CNCF, IBM Research, Dify, and more
- [For Open Source](https://dosu.dev/oss): Free tier for public repositories and OSS partner programs
- [Dosu vs Mintlify](https://dosu.dev/articles/dosu-vs-mintlify): How Dosu compares to Mintlify: no migration, internal engineering knowledge, and multi-repo coverage
- [Dosu vs Glean](https://dosu.dev/articles/dosu-vs-glean): How Dosu compares to Glean: vertical engineering knowledge that Dosu generates and maintains, alongside horizontal enterprise search
- [Dosu vs Unblocked](https://dosu.dev/articles/dosu-vs-unblocked): How Dosu compares to Unblocked: durable documents and a maintained knowledge layer rather than answer-only retrieval
- [Best Tools For AI Context](https://dosu.dev/articles/best-tools-for-ai-context): How DeepWiki, Context7, and GitBook compare to Dosu for agent context: multi-source ingestion, compounding knowledge, and lower agent token spend
- [Webinars](https://dosu.dev/webinars): Webinar recordings and transcripts from the Dosu team
- [Developer Resources](https://dosu.dev/developers): Index of developer resources, API docs, MCP server, and CLI

## Case Studies

- [Customers](https://dosu.dev/customers): Customer stories and case studies
- [Bluefin: Community Support After StackExchange](https://dosu.dev/customers/bluefin-community-support-after-stackexchange): Bluefin cut support response times from 42 hours to 12 minutes by using Dosu across six repositories.
- [How Dosu Powers CNCF’s Community Knowledge Platform](https://dosu.dev/customers/cncf-dosu-case-study): CNCF launched ask.cncf.io with Dosu at KubeCon NA 2025, answering 765 questions to help attendees navigate 200+ cloud native projects.
- [How IBM Research’s Docling Manages Open Source Growth with Dosu](https://dosu.dev/customers/docling-open-source-growth-with-dosu): IBM's Docling scaled from 0 to 40K GitHub stars (as of October 2025) in one year with just 4 maintainers. Check out how Dosu reduced response time by 99% and automated 70% of issues.
- [How TriliumNext Revitalized an Abandoned Open Source Project with Dosu’s Help](https://dosu.dev/customers/how-triliumnext-revitalized-an-abandoned-open-source-project-with-dosus-help): In the world of open source software, few stories are as remarkable as the resurrection of Trilium Notes. What began as a community fork of an abandoned project has evolved into a powerful collaboration.
- [Scaling support and documentation at Dify, one of the fastest growing OSS projects](https://dosu.dev/customers/scaling-support-and-documentation-at-Dify-one-of-the-fastest-growing-oss-projects): Learn how Dify uses Dosu to effortlessly manage hundreds of issues per month with a small team.
- [How Preset Keeps Apache Superset Healthy With AI-Driven Issue Triage](https://dosu.dev/customers/how-preset-keeps-apache-superset-healthy): Discover how Preset uses Dosu’s AI to automate GitHub issue triage for Apache Superset, reducing maintainer workload and speeding up community support.

## Articles

- [Articles](https://dosu.dev/articles): Guides and comparisons on agent memory, knowledge management, and AI context
- [The Four Costs of Coding Agent Memory and How to Measure Them](https://dosu.dev/articles/agent-memory-overhead): Agent memory costs you in four places. Measure the lookup, the tokens it adds to requests, capturing what sessions learn, and the review that keeps it true.
- [How to Set Up Coding Agent Memory: Local Files, Repo, or Shared Service?](https://dosu.dev/articles/where-should-agent-memory-live): Coding agent memory can live in three places. Choose one for each fact by who must approve it, whether it must outlive one laptop, and how often it changes.
- [Memory OS for AI Agents: What It Is and Which Platforms Offer One](https://dosu.dev/articles/memory-os-for-ai-agents): Of six agent memory projects, only MemOS calls itself an operating system. The label is not a capability, so compare these platforms on what each one does when a stored fact turns out to be wrong.
- [Agents + Smash at the Dosu Office: A Super Smash Bros Night for AI Builders](https://dosu.dev/articles/agents-smash-recap): We hosted 50+ people at our San Francisco office for Super Smash Bros, agent chat, and a few throwback consoles. Some AI engineers told us this was the only SF Tech Week event they planned to attend.
- [Enterprise Knowledge Management Systems With Native Connectors](https://dosu.dev/articles/enterprise-knowledge-management-connectors): Native connectors, custom connectors, and live lookups keep knowledge current in very different ways. Compare Dosu, Glean, Guru, Notion AI, and Confluence with Rovo on who maintains the integration, how long the staleness window runs, and whether source permissions survive the trip into search.
- [What Happens to Agent Memory When an Engineer Leaves](https://dosu.dev/articles/agent-memory-when-engineer-leaves): Some of what your coding agents know is committed to the repository and survives any departure. Some of it sits in a home directory on one laptop. Here is how to tell the layers apart before the machine is wiped.
- [Episodic vs Semantic vs Procedural Memory in AI Agents](https://dosu.dev/articles/episodic-vs-semantic-vs-procedural-memory): Three memory types, borrowed from two different cognitive-psychology taxonomies. What each one holds, what writes it, how it goes stale, and how to decide which store a new observation belongs in.
- [Checkmate at the Dosu Office: A Chess Night for AI Builders](https://dosu.dev/articles/chess-night-recap): We hosted 30+ developers at our San Francisco office for a blitz chess tournament. Players ranged from a titled FIDE Master to people learning the game for the first time.
- [Vector Database vs Knowledge Graph vs Files for Agent Memory](https://dosu.dev/articles/vector-database-vs-knowledge-graph-agent-memory): The substrate decides how a memory comes back, not whether it is still true. Compare vector stores, knowledge graphs, and version-controlled files on the write decision, retrieval, and what happens the day a stored fact stops being true.
- [How to Prune Stale Agent Memory](https://dosu.dev/articles/prune-stale-agent-memory): A fact your agent wrote down six months ago outlives the decision that produced it, and nothing in a normal session rechecks it. Here is how to tell which memories to retire, and how to fire that retirement off the change that made them false.
- [Why Coding Agents Forget Between Sessions and How to Fix It](https://dosu.dev/articles/coding-agents-forget-between-sessions): Losing the thread mid-task and starting cold in a new chat are two different failures with two different fixes. Here is how to tell which one you have before you change anything.
- [How to Onboard a New Engineer's Coding Agent](https://dosu.dev/articles/onboard-new-engineer-coding-agent): A tenured engineer has spent months correcting their agent, and most of that correction never reached the repository. Here is which context a new hire's agent loads on day one, what stays behind on the old laptop, and how to check which one you have.
- [Best MCP Memory Servers for Shared Team Agent Memory](https://dosu.dev/articles/best-mcp-memory-servers): Where an MCP memory server keeps its store, on one laptop or somewhere the whole team reads, matters more than how it extracts or retrieves. Compare seven by where the store lives, who can reach it, and what stands between one agent writing a fact and every other agent retrieving it.
- [Agent Memory Governance for Engineering Teams](https://dosu.dev/articles/agent-memory-governance): Once agents write to shared memory, one wrong fact becomes something every agent repeats. Six controls decide whether you catch it: what gets written, where it came from, who can read it, how you correct it, which version wins, and what gets deleted.
- [Team Memory for Coding Agents: How to Share Context Across a Dev Team](https://dosu.dev/articles/team-memory-for-coding-agents): Coding agent memory is per developer and per machine by default, so what one engineer teaches their agent never reaches the rest of the team. Here is what each layer of shared context can carry, and how to set one up.
- [How to Make Coding Agents Remember Architectural Decisions](https://dosu.dev/articles/coding-agents-remember-decisions): Your repository records what the team decided, not why, so a coding agent keeps proposing the approach you rejected last spring. Here is where decision rationale lives and how to make it reachable.
- [Agent Memory vs RAG vs Bigger Context Windows](https://dosu.dev/articles/agent-memory-vs-rag-vs-context-window): Three mechanisms get called agent memory, and they fail differently. Compare what each one stores, when it runs, and which cold-start symptom it fixes.
- [Best AI Knowledge Bases for Engineering Teams in 2026](https://dosu.dev/articles/best-ai-knowledge-bases): Two questions separate these products: does the knowledge base build itself from your engineering activity, and can a coding agent query it mid-session. Here is where Dosu, Falconer, Glean, Unblocked, Cognee, and Mintlify land.
- [Best AI Agent Memory API Providers in 2026](https://dosu.dev/articles/agent-memory-api-providers): Compare Dosu, Mem0, Zep, Supermemory, Letta, LangMem, and Redis or Postgres by what each remembers, what triggers a write, and where repository context fits for coding agents.
- [How to Choose Context Engineering Tools for Claude Code in 2026](https://dosu.dev/articles/context-engineering-tools-claude-code): Compare Claude Code built-ins with tools for live documentation, repository retrieval, shared knowledge, and persistent memory.
- [Best Knowledge Management Systems for Startups in 2026](https://dosu.dev/articles/knowledge-management-for-startups): Compare seven knowledge management systems for startups by source coverage, setup effort, and how they help teams create, review, and maintain knowledge.
- [Centralized Knowledge Repository Tools for Distributed Engineering Teams](https://dosu.dev/articles/centralized-knowledge-repository-tools): Compare wikis, enterprise search, agent memory stores, and automatic capture for distributed engineering teams by how they capture knowledge, maintain it, and control access.
- [Best DeepWiki Alternatives for Engineering Teams](https://dosu.dev/articles/deepwiki-alternatives): Compare Dosu, Google Code Wiki, OpenWiki, Repowise, RepoWiki, deepwiki-open, and Swimm on keeping a repo wiki current, private code, and coding agents.
- [Best Tools to Keep AGENTS.md Files Up to Date](https://dosu.dev/articles/tools-to-keep-agents-md-updated): Your AGENTS.md is a snapshot and your agents act on it every session. Compare Dosu, Falconer, Unblocked, the open source drift checkers, memory layers, and docs platforms on what actually triggers an update.
- [Knowledge Base Software for Slack, GitHub, and Jira](https://dosu.dev/articles/knowledge-base-software-slack-github-jira): Dosu, Glean, Guru, Slite, Confluence, Unblocked, Notion AI, and Nuclino compared on what each reads from Slack, GitHub, and Jira and how docs stay current.
- [The Best AI Knowledge Management Systems for Search Accuracy in 2026](https://dosu.dev/articles/best-ai-knowledge-management-search-accuracy): Accuracy percentages in vendor decks are unfalsifiable. Retrieval architecture is not. Here is what to ask about hybrid search, reranking, chunking, and freshness, and how Dosu, Unblocked, Mintlify, DeepWiki, and Notion differ.
- [How To Connect Confluence Docs to LLMs and Coding Agents](https://dosu.dev/articles/connect-confluence-docs-to-llms): Your architecture decisions are in Confluence and your coding agents cannot read them. Compare Dosu, Atlassian Rovo, Unblocked, and Mintlify on indexing coverage, sync latency, and whether an agent can write back.
- [Best AI Knowledge Management Tools for Engineering Teams in 2026](https://dosu.dev/articles/best-ai-knowledge-management-tools): Most AI knowledge tools help people find what a colleague wrote. Coding agents need current context. Compare Dosu, Unblocked, Glean, Confluence, Notion AI, Guru, and more.
- [Best Automatic Knowledge Capture Tools for AI Coding Agents](https://dosu.dev/articles/automatic-knowledge-capture-tools): Capture is only half the problem. Compare Dosu, Falconer, Glean, Guru, and Notion AI by what triggers an update, who approves it, and whether the result ever reaches a coding agent.
- [Best AI Agent Memory Tools in 2026](https://dosu.dev/articles/best-ai-agent-memory-tools-2026): Compare Mem0, Zep, Letta, Cognee, LightMem, and Dosu by use case, capture method, maintenance model, and retrieval controls for AI agent memory.

## Webinars

- [Webinars](https://dosu.dev/webinars): Webinar recordings with full transcripts
- [AI Memory News: Muse Wonder, Haiku 5.5 and Decision Models](https://dosu.dev/webinars/ai-memory-news-muse-wonder-haiku-5-5-decision-models): Taylor Dolezal, Devin Stein and Michael Mangus on Muse Wonder, Haiku 5.5, Cognition's memory, decision models and AI-written tests.
- [Ranking AI Systems: AI Memory Benchmarks](https://dosu.dev/webinars/ranking-ai-systems-ai-memory-benchmarks): Taylor Dolezal, Devin Stein and Michael Mangus rank seven agent memory benchmarks, from LoCoMo to VibeMemBench, by how much they tell coding agent teams.
- [Agent Memory: What Happens When It's Wrong?](https://dosu.dev/webinars/agent-memory-what-happens-when-its-wrong): Taylor Dolezal, Devin Stein and Michael Mangus on keeping agent memory current: decay, dreaming, external events, what it costs, and how to tell if it works.
- [Agent Memory: How Does It Find Anything?](https://dosu.dev/webinars/agent-memory-how-does-it-find-anything): Taylor Dolezal and Michael Mangus on the read side of agent memory: why agents retrieve, what the harness wants, how search evolved, and why selectivity wins.
- [Agent Memory: What Information Is Worth Remembering?](https://dosu.dev/webinars/agent-memory-what-information-is-worth-remembering): Taylor Dolezal, Michael Mangus, and Devin Stein on the write side of agent memory: what's worth storing, the tests before writing, schema, and org-scale knowledge.
- [Agent Memory: Where Does Knowledge Live?](https://dosu.dev/webinars/agent-memory-where-does-knowledge-live): Taylor Dolezal and Devin Stein on where agent memory lives: token sequences, graphs and trees, storage tiers, and why most teams should start with Git.
- [Agent Memory: What Are the Building Blocks of a Memory System?](https://dosu.dev/webinars/agent-memory-building-blocks): Devin Stein, Michael Mangus, and Taylor Dolezal break down a survey paper on agent memory: representation, extraction, retrieval, and maintenance.
- [LLM Wikis and How to Give Your Agents Memory](https://dosu.dev/webinars/llm-wikis-and-how-to-give-your-agents-memory): Harrison Chase moderates Brace Sproul, Devin Stein, and Jeff Huber on LLM wikis: what they are, whether they're the right abstraction, and how to measure ROI.
- [Your AI Coding Assistant Doesn't Know Your Project. Let's Fix That!](https://dosu.dev/webinars/your-ai-coding-assistant-doesnt-know-your-project): Dosu founder Devin Stein and DBOS's Alex Poliakov on how agents share knowledge through the Dosu MCP, and why durable execution powers Dosu's AI workflows.

## Blog

- [Blog](https://dosu.dev/blog): Insights on AI documentation, knowledge management, and engineering workflows
- [Full blog index (Markdown)](https://dosu.dev/blog.md): Every published post, newest first, linking each post's Markdown version
- [Turning Coding Sessions Into Shared Agent Memory](https://dosu.dev/blog/coding-agent-memory-from-session-logs): Recurring discoveries in Claude Code and Codex show why shared agent memory needs explanations and sources.
- [How 20 Vercel AI SDK tickets drove the agent cost gap](https://dosu.dev/blog/claude-code-vs-codex-token-costs): Claude Code on Opus 5.5 and Codex on GPT 6.1 Sol attempted the same 49 tickets. The recorded cost gap came entirely from high-scope work.
- [Procedural Memory for Coding Agents: Turning Knowledge Into Action](https://dosu.dev/blog/procedural-memory-for-coding-agents): Learn how to turn a coding agent's discovery into a shared procedure, choose where the instructions belong, and check that the next agent uses them correctly.
- [Semantic memory for coding agents needs a history](https://dosu.dev/blog/semantic-memory-for-coding-agents): Your coding agent needs to know which facts still apply. Our Celery-to-DBOS migration shows why semantic memory needs scope, dates, and source evidence.
- [Episodic memory for coding agents: What survives your sessions](https://dosu.dev/blog/episodic-memory-for-coding-agents): Episodic memory is a coding agent's record of past sessions. Saved lessons can outlive the transcripts behind them, so keep each lesson tied to its source.
- [September Drop: Dosu goes back to school](https://dosu.dev/blog/september-2026-dosu-drop): Dosu studies coding sessions, shows what your agent reads, and makes your credits go further.
- [How to Build Agent Memory: Maintenance](https://dosu.dev/blog/how-to-build-agent-memory-maintenance): How to keep agent memory useful as things change: preserve evidence, scope updates, choose maintenance triggers, and measure whether memories still help.
- [How to Build Agent Memory: Search & Retrieval](https://dosu.dev/blog/how-to-build-agent-memory-search-and-retrieval): How agents find and use what they've learned: choosing a search strategy, returning useful context, and balancing precise evidence with higher-level understanding.
- [A Correct Translation Can Still Be Wrong in Meaning](https://dosu.dev/blog/context-for-localization): A settings label translated 'correctly' and still meant the wrong thing. The fix wasn't a better dictionary, but context — and it's already scattered across your code, PRs, and conversations.
- [How to Build Agent Memory: Knowledge Capture](https://dosu.dev/blog/how-to-build-agent-memory-knowledge-capture): How to decide what agents should remember: five questions for evaluating knowledge, choosing its structure, and capturing discoveries for future work.
- [How to design a CLI login flow for coding agents](https://dosu.dev/blog/cli-login-flow-for-coding-agents): Design a CLI login for coding agents with a ticket flow that pauses, resumes in a new process, and returns the next command in JSON. Learn how Dosu built it.
- [How to Build Agent Memory: Where Does Knowledge Live?](https://dosu.dev/blog/how-to-build-agent-memory-where-does-knowledge-live): Agent memory starts with representation and storage. Here's how to decide what knowledge looks like, where it lives, and what is worth maintaining.
- [Your coding agent budget pays for context, not code](https://dosu.dev/blog/agent-budgets-pay-for-context-not-code): Planning and gathering context account for two thirds of agent costs.
- [OSS has Changed. So has Dosu.](https://dosu.dev/blog/oss-has-changed-so-has-dosu): Changes coming to Dosu OSS plan on Sept 1st. ~96% of OSS projects see no change.
- [How to Build Agent Memory: The Building Blocks of Agent Memory](https://dosu.dev/blog/the-building-blocks-of-agent-memory): We break down 'Are We Ready for an Agent-Native Memory System?' — the four building blocks of agent memory: storage, extraction, retrieval, and maintenance.

## Company

- [Careers](https://dosu.dev/careers): Join the team building the future of documentation
- [Contact](https://dosu.dev/contact): Contact support, sales, partnerships, privacy, and the Dosu office
- [Legal — Terms of Service](https://dosu.dev/legal/terms): Terms of service
- [Legal — Privacy Policy](https://dosu.dev/legal/privacy): Privacy policy

## Optional

- [Complete index (llms-full.txt)](https://dosu.dev/llms-full.txt): This index expanded with every blog post, product details, and FAQs
- [Blog Topic: Product Updates](https://dosu.dev/blog/topics/product-updates): Monthly Dosu Drops and launch posts on what's new in Dosu, from Libraries and Agents to Templates and Review.
- [Blog Topic: Documentation](https://dosu.dev/blog/topics/documentation): How to keep docs accurate as code changes: drift detection, freshness checks, and the case for writing things down.
- [Blog Topic: Agent Memory](https://dosu.dev/blog/topics/agent-memory): How coding agents store, find, and maintain the context they need, from the How to Build Agent Memory series to research on what agents read.
- [Blog Topic: Engineering](https://dosu.dev/blog/topics/engineering): How we build Dosu: pipelines, evals, infrastructure, and lessons from running AI agents in production.
- [Blog Topic: Open Source](https://dosu.dev/blog/topics/open-source): Issue triage, labeling, and maintainer workload in open source projects, plus the tools we release for maintainers.
- [Blog Topic: Case Studies](https://dosu.dev/blog/topics/case-studies): How teams and open source communities use Dosu, with the numbers behind it.
- [Blog Tags](https://dosu.dev/blog/tags/agents): Browse posts by tag, starting with agents; every post links its own tags
- [Blog Authors](https://dosu.dev/blog/authors): Everyone who writes for the Dosu blog, with links to their posts