Office docs to Markdown in milliseconds, plus YC's fleet-of-agents harness
Every AI project hits the same wall in week one: the data you need is trapped in Word docs, PDFs, and somebody's Excel export. anydoc, this issue's featured tool, is Firecrawl's converter for exactly that. Word, PowerPoint, Excel, PDF, EPUB, all of it comes out as clean Markdown in milliseconds, and it's the fastest-moving project we're tracking right now. One catch worth knowing up front: it reads the text layer of a PDF, not scans. If your PDFs are photographs of paper, you still need OCR first. The rest of the issue is agents growing up. Comp AI built a CRM where the agent does the work on its own schedule and the CRM is just where it takes notes. qm is Y Combinator's own system for running a whole team's agents, one isolated workspace per person, driven from Slack or the web. The pattern is hard to miss: agents are moving out of demos and into org charts. And because somebody has to keep the lights on, Authentik rounds it out: self-hosted single sign-on and multi-factor auth, the stuff you'd otherwise buy from Okta or Auth0. Less exciting than the agent stuff, and probably the one your company actually needs first.
Convert Word, PowerPoint, Excel, OpenDocument, RTF, EPUB, CSV, and PDF to clean Markdown. Built in Rust, with Node.js and Python bindings.
The Lens
anydoc turns Word, PowerPoint, Excel, OpenDocument, RTF, EPUB, CSV and PDF files into clean Markdown, the format LLMs read best. Every format lands in one shared document model and one Markdown writer, so a spreadsheet table and a Word table come out the same shape. Firecrawl built it, and it's MIT licensed and free.
It's a library, not a server. The core is Rust, with packages for Node, Python and the browser through WebAssembly, plus a CLI: npx @firecrawl/anydoc report.docx. No API key, nothing to host, and a reported median under 5 ms per document.
The paywall is OCR. anydoc does none, so a scanned PDF fails with NeedsOcr. Opt in with --ocr hosted and that file goes to Firecrawl Parse at 1 credit per page: 1,000 free credits a month, then Hobby at $19/month. Solo and small teams: free. Teams buried in scans: budget for Parse or bring your own OCR.
The catch: the benchmark that makes it look great is Firecrawl's own, judged by an LLM on a private corpus, with no PDF row in the per-format results. Hosted OCR also ships the whole document to Firecrawl, not just the scanned pages. Test your ugliest real files first.
An open-source, agentic-first CRM.
The Lens
Comp AI CRM flips the usual setup: an AI agent runs on its own schedule and the CRM is where it keeps notes. It reads your team's inbox, turns new people into contacts, enriches companies and books its own follow-ups, with an Agent tab on each record showing its reasoning. It's MIT, from Comp AI, the compliance-software company.
Self-hosting means three deployments plus Postgres: a Next.js app, a NestJS API and the agent. You need Bun, Docker and a Google or Microsoft OAuth app, and it's single-tenant by design. It leans hard on Vercel: the model runs through Vercel's AI Gateway, keyless only when deployed on Vercel.
No paid tier, no hosted version. The model calls cost money, though: AI Gateway bills provider list prices past a $5 monthly credit, and LinkedIn and company data come from Context, free for 1,000 credits a month, then $19/month. Solo and small sales teams: free software, budget for tokens. Bigger orgs: this isn't Salesforce.
The catch: it's young and has gone quiet. The last release shipped August 21 and the main branches haven't moved since. It also sends anonymous telemetry daily unless you set CRM_TELEMETRY_DISABLED=1. Test it on copied data first.
Flexible identity provider
The Lens
authentik is a self-hosted identity provider: one login (SSO, single sign-on) for every app your team uses, plus MFA, a user directory, and LDAP for older gear. It speaks SAML, OIDC and SCIM, and the open source edition is MIT-licensed and free. It's the self-hosted answer to Okta, Auth0 and Entra ID. Setup is Docker Compose with three containers: Postgres, a server and a worker. No Redis since 2025.10; the minimum is 2 CPU cores and 2 GB of RAM. The time sink is flows: every login, signup and recovery path is a chain of stages you wire yourself. Powerful, not intuitive. Enterprise is $5 per user per month plus $0.02 per external user, billed annually, for Entra ID and Google Workspace sync, smart-card login and compliance audit logs. Ticket support starts above $1k a year. Solo and small teams: free. Teams syncing Microsoft or Google: Enterprise. Large orgs: Enterprise Plus from $20k a year. The catch: you own the patching. authentik published two dozen security advisories in 2026, including a critical code-execution bug and five SAML flaws that let an attacker sign in as someone else. Fixes ship fast. An unpatched identity provider is an open front door.
Multiplayer agent harness for work
The Lens
qm is Y Combinator's own open source answer to running a fleet of AI agents across a team. It's a "multiplayer agent harness for work": every person gets an isolated workspace with personal and shared scopes, you drive it from Slack or a web UI, and it can run background and scheduled agent jobs. This is YC Labs code, not a fan project trading on the name, the release is announced on ycombinator.com and served from qm.ycombinator.com. The interesting part is the multiplayer framing. Instead of each developer babysitting one agent in a terminal, qm treats agents as shared, always-on coworkers with their own scopes and cron schedules. For a team already leaning on coding agents, it's a free way to run them coordinated instead of siloed. Solo use works too, but the design clearly targets groups. The catch is that it's brand new. The commit history is short and the stars arrived fast, which means the code is early and production-readiness is unproven. The pedigree is real and the idea is sharp, but pilot it on non-critical work before you wire it into how your team actually ships.
Get the next issue in your inbox
Free. No spam. Unsubscribe anytime.