All agents

Almanacreview, pricing and limits

by Almanac

The AI agent with its own computer and a self-updating company wiki

generalcoding
checked
Price
$30–$200/mo
Free tier
No
Autonomy
Semi-autonomous
Stage
Beta

Almanac replaces a chatbot bolted onto separate integrations for founders and ops teams at early-stage companies: it already reads across nine everyday tools, from Gmail to GitHub, before it starts a task. You text it work over Slack or iMessage, and it works from that memory instead of asking you to re-explain context, then tells you when it's done.

Almanac is a semi-autonomous AI agent, launched from Y Combinator's Summer 2026 batch, that runs on its own computer with a browser and terminal. It connects Gmail, Slack, GitHub, Attio and Calendar into a self-updating company wiki, then acts on tasks while pausing at logins, payments, or decisions needing human approval.

Maker: Almanac · Autonomy: semi autonomous · Maturity: BETA

About Almanac

Almanac is an AI agent from Kushagra Chitkara, Rohan Sharma, and Divit Sheth, who previously built agent-context wikis for Harvard and NASA before founding the company out of Y Combinator's Summer 2026 batch in San Francisco. Their pitch: an agent with a company brain, not another chatbot wired into a few integrations. The product centers on a wiki that stays current on its own. Every connected account, from email and calendars to code repositories and CRM records, gets folded into pages that inform each new request, and those same pages are shared across every teammate who has connected an account. Rather than starting a fresh context window each session, Almanac starts from what the company already knows, then decides on its own how far to take a task before checking back in with a person. Founders describe handing it recurring operational work: sorting through customer feedback channels, keeping a renewal account's history current for a sales conversation, watching a production environment for small breakages, and managing personal admin like unused subscriptions. It targets small teams still figuring out who owns what, rather than an enterprise with dedicated tooling for each function already in place. Access is entirely conversational: message it over Slack or iMessage rather than opening a dashboard, though the underlying wiki is also readable on the web. Every plan opens with a trial period before billing begins, and pricing scales by how much model usage and browser time a seat consumes each month, up to a custom Enterprise tier for larger rollouts. Because the product only recently launched publicly, there is not yet a public changelog or independent benchmark to point to. The founders' background, building wiki-style memory systems for research institutions before turning the same idea toward companies, is the clearest signal of where the product is likely headed next.

Pricing

7-day trial, then $30/seat/month (Starter), $100/seat/month (Pro, most popular, 5x usage / 2x browser minutes), $200/seat/month (Scale, 20x usage / 4x browser minutes), or custom Enterprise pricing with SSO and access controls. Teammate invites are free until accepted.

Key Features

  • Self-updating company wiki: Compiles Gmail, Google Calendar, Google Docs, GitHub, Attio, Granola meeting notes and past AI conversations into wiki pages it reads before acting on any task.
  • Its own computer, browser and terminal: Runs on a dedicated machine with a real browser and terminal, so it can sign into tools that have no API integration the same way a person would.
  • Escalation at the right boundary: Pauses and pings the user on Slack or iMessage, or hands over a live browser session, at any login, payment, or decision it judges it should not make alone.
  • Shared team memory: Every teammate's Almanac reads and writes the same company wiki, so a question answered for one person stays answered for the rest of the team.
  • Context for other coding agents: Exposes the same wiki to agents like Claude Code and Codex, so a coding session starts already caught up on the company instead of re-deriving context.
  • Automations written as sentences: Recurring work is defined in one plain sentence instead of a workflow builder, for example telling it to reproduce and file every new customer bug report.

Strengths

  • Runs on its own computer with a working browser and terminal, so it can operate accounts like GitHub, Attio, and a connected bank that have no clean API for a chatbot to call.
  • Explicit escalation boundary: it stops itself before any sign-in, purchase, or judgment call and either notifies the user or opens a live browser session for them to take over.
  • Founders spent a year and a half building agent wikis for Harvard and NASA before Almanac, which shows in a persistent memory layer most new company-assistant startups have not built yet.

Weaknesses

  • No SWE-bench, WebArena, or GAIA benchmark has been published for Almanac, so there is no independent, third-party measure of its task success rate yet.
  • There is no permanent free tier, only a 7-day trial before billing starts, which rules out casual or personal use compared to free chatbot-based assistants.
  • Being a brand-new product, Almanac has no long public track record with paying customers yet, and no public API or CLI is documented for developers who want to build on it.

Frequently Asked Questions

How much does Almanac cost in 2026?

Almanac has four pricing tiers. Starter costs $30 per user each month, Pro (most popular) costs $100 with five times the model usage and double the browser minutes, and Scale costs $200 with twenty times the model usage and four times the browser minutes of Starter. The Enterprise tier is negotiated individually and includes single sign-on and rollout controls, and every tier begins with a 7-day trial.

Is Almanac fully autonomous?

Almanac is semi-autonomous, not just a scripted chatbot: founders describe it independently triaging a support channel, filing GitHub issues, and drafting a full renewal deck without step-by-step direction. It only stops itself at higher-stakes boundaries, such as an account sign-in, a purchase, or a judgment call outside its risk tolerance, and hands over a live browser session for a person to finish those parts.

What are the best alternatives to Almanac?

Teams that want the underlying agent framework rather than a hosted product often look at Hermes Agent, the open-source project Almanac's own pitch nods to as inspiration; running it yourself means no monthly seat fee but no shared company wiki either. OpenClaw is the pick for a single developer who wants a capable local agent without any subscription. For a narrower brief, mainly writing and shipping code changes, Claude Code stays closer to the terminal than Almanac's broader operational scope.

How does Almanac compare to Hermes Agent in 2026?

Both build a memory layer that improves the more it is used, but Hermes Agent is open source and self-hosted while Almanac is a hosted product with no source code released. Almanac's differentiator is a shared company wiki that every teammate's instance reads and writes, plus its own computer with a browser and terminal, versus Hermes Agent's per-user workflow memory. Choose Almanac when several people need one shared source of truth; choose Hermes Agent to self-host and keep data in-house.

How do you get started with Almanac?

Sign up at usealmanac.com and connect the accounts you want it to use, such as Gmail, Google Calendar, Slack, or GitHub; invitations to teammates don't count against the bill until accepted. From there, message it a task over Slack or iMessage, for example asking it to triage a support channel or draft a meeting brief, and it works from the wiki it has already built from your connected tools. The first week runs on the trial before any plan is billed.

Top Alternatives

  • Hermes Agent: Pick Hermes Agent if you want to self-host an open-source workflow-learning agent instead of Almanac's hosted, company-wide wiki.
  • OpenClaw: Pick OpenClaw if you want a free, local-first agent for coding and browser tasks without a monthly per-seat cost.
  • Claude Code: Pick Claude Code if you mainly need an agent shipping pull requests from the terminal, not one managing company-wide operational memory.

HokAI guides covering Almanac

More AI Agents on HokAI

Visit Almanac Official Site