Backed byY Combinator

Your team’s AI work,
as context every agent can use.

Nessie ingests every session across ChatGPT, Claude, Codex, Cursor, and 20+ more tools, then serves it back to your people and their agents: searchable, permissioned, and grounded in the original sources.

Book a demo
Download for Mac
A hand tucking a handwritten note into an open leather journal on a wooden desk
Used by people and teams at
Basic Capital
OMNIA Partners
TD Ameritrade
Alpha School
Y Combinator
01Sources

Connect once.
Backfill everything.

Connect an integration and Nessie backfills all available history from it, then keeps new sessions, meetings, and files flowing in. No copy-pasting, no browser extension, and multiple accounts per integration.

  • AntigravityAgent
  • Apple NotesNotes
  • ChatGPTChat
  • ClaudeChat
  • Claude CodeAgent
  • Claude CoworkAgent
  • CodexAgent
  • CursorAgent
  • GeminiChat
  • GmailMail
  • GranolaMeetings
  • GrokChat
  • Grok BotAgent
  • HermesAgent
  • iMessageChat
  • Local FolderFiles
  • ManusAgent
  • Microsoft TeamsMeetings
  • ObsidianNotes
  • OpenClawAgent
  • OpenCodeAgent
  • OutlookMail
  • PerplexityChat
  • PiAgent
  • PlaudMeetings
  • Read AIMeetings
  • SharePointFiles
  • More every releaseChangelog →
02Agent access

Every agent starts
where the last one stopped.

Your agents query Nessie directly over CLI, MCP, and skills. They search past sessions, read full transcripts, and pick up a teammate’s work with the decisions and blockers intact, without anyone copying context between tools.

Codex

Nessie Pick up my Claude Code session about the marketing site server-side rendering. Where did we leave off?

Nessie Found your Claude Code SSR session from Tuesday
Nessie Read its goal, files, decisions, and remaining work

I found your earlier Claude Code session and checked it against the current branch. The SSR migration itself is complete:

  • PR #389 merged and deploy 142 is healthy; the main landing, pricing, and docs routes now render metadata on the first response.
  • Route tests and the production smoke check passed after the hosting cutover.
  • One follow-up remains: preview pages still fetch metadata client-side, so shared-link unfurls can be stale.

Next step: Move preview metadata into the SSR loader, add a route test, then verify one shared preview against the local server.

Message your agent

One contract across every surface.

The CLI, MCP server, and agent skills expose the same commands with the same permissions. An agent sees exactly what the person running it can see, and nothing more.

Read the agent docs 
$ nessie ls --type transcript --name billing
nessie ls: showing 1-8 of 8
kind              owner           updated       id                                    name
claude_code_chat  maya@acme.com   Oct 14 16:08  8f2c41d0-9b7e-4c3a-a5d1-2e6f7a8b9c01  Annual-plan proration
codex_chat        me              Oct 13 11:42  2d7b90e4-51c3-4f8a-b6e2-9a0c1d3e5f47  Metered rollback plan
claude_code_chat  priya@acme.com  Oct 12 09:15  5a1e77c3-08d2-4b6f-9e3a-c4d5e6f7a8b9  Subscription backfill
cursor_chat       me              Oct 11 17:30  c3d4e5f6-7a8b-4c9d-8e0f-1a2b3c4d5e6f  Webhook retry handler
chatgpt_chat      sam@acme.com    Oct 10 14:02  9b8a7c6d-5e4f-4a3b-9c2d-1e0f2a3b4c5d  Stripe metered pricing
claude_chat       me              Oct 09 10:21  4e5f6a7b-8c9d-4e0f-a1b2-c3d4e5f6a7b8  Billing FAQ for support
codex_chat        maya@acme.com   Oct 08 16:47  7d6c5b4a-3e2f-4d1c-b0a9-8f7e6d5c4b3a  Invoice PDF regression
opencode_chat     priya@acme.com  Oct 07 12:05  1f2e3d4c-5b6a-4978-8a9b-0c1d2e3f4a5b  Tax rounding fix
03Access control

Private by default.
Shared on purpose.

Context only moves when someone decides it should. Every share has an owner, a scope, and an audience, and agents inherit exactly the permissions of the person running them.

  • Scoped sharingShare one session, every session in a repository, a whole integration, or a context. Everything else stays private.
  • People, teams, and adminsGrant access to a person, a team, or admins only. Change or revoke it at any time.
  • Privacy Mode by defaultNo one at Nessie can read your content, and it is never used to train models.
  • Local-firstYour library starts on your machine. Cloud Sync is opt-in for teams, MCP, and remote agents.
  • On-prem deploymentRun the whole system on your own infrastructure when your organization requires it.

Give a person or team access to one Claude Code session without exposing the rest of your work.

Security practices and policies are published in our Trust Center.Visit Trust Center
04FAQ

Questions teams
ask before rollout.

What is Nessie?

Nessie is the context layer for AI-native work. It ingests all available history from the AI conversations and local sources you already use, then turns them into context you, your team, and your agents can reuse.

Is Nessie another AI chat app?

No. You keep using ChatGPT, Claude, Claude Code, Codex, OpenCode, Cursor, and the tools you already work in. Nessie sits above them as the shared memory and context layer.

What does Nessie sync?

Nessie ingests all available history from AI conversations, local agent sessions, Obsidian vaults, Granola meeting notes, Read AI meeting reports, and other integrations, then keeps new activity synced. It also tracks generated contexts, profile signals, and source metadata. You can connect multiple accounts or source roots for the same integration.

Can I use Nessie as an individual?

Yes. Nessie works just as well for one person as for a team - it unifies your own AI conversations and agent traces across every tool and makes them searchable by you and your agents. Team features like shared context are there when you need them.

Where is my data stored?

Nessie is local-first by default. Your library starts on your Mac, and Cloud Sync is optional for MCP, teams, multiple devices, and remote agent access. To generate summaries, search, and context, Nessie processes your conversations with third-party AI providers in the moment, but it never stores that content on our servers or uses it to train models.

Can anyone at Nessie see my conversations?

By default, no. Whether anyone at Nessie can read your content comes down to one setting: in Privacy Mode (the default) no one can, and your content is never used to improve our systems. If you opt into Help Improve Nessie, you choose to share some transcript snippets and usage data with us. Cloud Sync is a separate, storage-only setting: it stores your library on our servers (encrypted) so it syncs across devices, but even with Cloud Sync on, in Privacy Mode we will not look at your conversations. See our Privacy Policy for the details.

How do agents use Nessie?

Agents use Nessie through CLI and MCP. They can search your history, read full transcripts, run check-ins, resume prior sessions, and generate contexts from source conversations.

Do I need a Chrome extension?

No. Nessie ingests all available history from connected accounts, local app data, agent traces, and vault folders without asking you to copy-paste conversations into a browser extension.

Can I deploy Nessie on-prem?

Nessie supports on-prem deployments for organizations that need to keep the whole system on their own infrastructure. Talk to us about an on-prem deployment.

05Demo

See Nessie on
your team’s real work.

A walkthrough of how ingestion, sharing, and agent access fit your stack, and what a rollout looks like for your team.