Share conversation context from Claude.ai to Claude Code

This is critical for anyone who uses both claude.ai and Claude Code.

Current workflow is broken:

1. Plan/brainstorm in claude.ai (has memory, good for exploration)
2. Ready to implement → switch to Claude Code
3. Claude Code has zero context from the planning phase
4. Manually re-explain everything or copy-paste walls of text
5. Lose nuance, waste time, break flow

What should happen:

claude --from-web <conversation-id>
# or
claude --with-context "my project planning session"

Claude Code starts with full context from the web conversation.

Implementation options:

1. Direct integration - Claude Code can query claude.ai conversation API
2. Export/import - claude.ai exports context file, Claude Code imports
3. Shared backend - Both products read from same context store

If none of these are feasible short-term:

Update ToS to permit subscribers to build their own bridges. We'll extract our own context and feed it to Claude Code. Just don't ban us for accessing our own data.

The web UI has memory but is unstable. Claude Code is stable but has no memory. Let us combine the best of both.

+1

Definitely needed!

Hello Claude Team,

Please do NOT implement export/import (results in silos) — need only live read access to chat history and artifacts for always current and single source of truth.

The inability of Claude Code to connect to the same database that Claude.ai connects to using same user creds is a HUGE shortcoming and frustration.

Ten years from now, we will be joking to our kids, “Back in my day, I paid $200/month for Max 20x, had 60+ conversations with detailed architecture specs, database schemas, biz models, security patterns... Claude Code Opus 4.5 in 2026 still did not integrate with our chats/artifacts on Claude.ai at all using same account creds and it was all manual wading through of long, maxed out, overwhelming chats and copy/pasting which is quite the opposite of what AI is supposed to help us with. We'd scroll through conversations so long the browser would lag. Copy-paste code blocks one at a time. Manually maintain CLAUDE.md files like some kind of digital scrapbook. And if you forgot which chat had that one critical decision about session token handling? Good luck — here's a search box that returns 50 results."

That's like what used to be printing out MapQuest directions or manually typing URLs from TV commercials.

All of that institutional knowledge exists right here in my conversations. And Claude Code can't see any of it without me playing human middleware that wastes too much precious time.

The irony:

Anthropic is building AI to reduce tedious manual work... while requiring tedious manual work to use their own products together.

The frustration:

"I'm literally building automation software while manually copying context between your tools."

Claude Code should have live read access to Claude.ai chat history and artifacts

Problem Statement

Claude Code and Claude.ai are completely disconnected despite:
Same authenticated user, Anthropic account credentials, same $200/month Max plan
Claude Code already connects to external services via MCP
Connecting to Anthropic's own conversation API should be trivial

When I open Claude Code, it has zero knowledge of the extensive architectural discussions, design decisions, code reviews, and artifacts I've created in Claude.ai. I'm paying for one Claude, but getting two strangers who don't talk to each other.

When I switch to Claude Code to implement these designs, I have to manually wade through long, maxed-out conversations, copy-paste context, and maintain CLAUDE.md files like a digital scrapbook. This is exactly the kind of tedious manual work AI is supposed to eliminate.

Proposed Solution:
Live read access — not export/import.

Claude Code already:

• Authenticates against the same Anthropic account
• Supports MCP connections to external services (Google Drive, Slack, Asana, etc.)

Connecting to Anthropic's own conversation API should be simpler than connecting to third-party services:

Expose chat history via internal MCP server or API endpoint
Claude Code can query: "What did we decide about session token handling?"
Live access — always current, single source of truth

Why NOT export/import (re: Issue #13843)

Export/import creates unnecessary data silos:

• Point-in-time snapshots that immediately become stale
• Manual sync overhead pushed onto the user
• Data duplication with no single source of truth
• Violates basic architectural principles we'd reject in any production system

I can already search my Claude.ai chat history from within Claude.ai using conversation_search and recent_chats tools. The capability exists — it's just walled off from Claude Code for no apparent technical reason.

The Ask

1. Give Claude Code read access to the authenticated user's Claude.ai conversation history
2. Allow searching by keywords, topics, date ranges (same as Claude.ai's existing tools)
3. Allow searching through and reading artifacts content from past conversations
4. No export files, no manual sync — live access to the same data

Environment

• Plan: Max ($100-200/month)
• Primary use: Enterprise application development
• Claude.ai usage: Heavy — architectural planning, design reviews, documentation
• Claude Code usage: Implementation of designs discussed in Claude.ai

Please do prioritize this severe pain-point issue and implement it in your NEXT release, if possible.

---

Related issues: #14228, #13843

I griped about this in web Help yesterday, and was informed "This design is intentional across all our platforms - each interface maintains separate conversation histories for privacy and organization purposes."

Really poor Ux, all the chats are in my account, how can there possibly be a permissions problem? And don't all the conversations store in the same data store/history? Pretty sure I've been throttled on the code side and had the throttle timeout appear on the web side... Why the silo'ing then?

+eleventy-five on this "feature"

built a tool for this if anyone wants to try it: https://github.com/selviarora/claude-context-bridge

chrome extension with a button on claude.ai. saves the conversation locally and exposes it to claude code via mcp

check it out :)

+1

built a tool for this if anyone wants to try it: https://github.com/selviarora/claude-context-bridge chrome extension with a button on claude.ai. saves the conversation locally and exposes it to claude code via mcp check it out :)

tried exact the same as the tutorial says.

"Specified native messaging host not found."

+1

+1

this is my biggest friction point too. I plan features in Claude.ai with artifacts and diagrams, then have to manually paste the decisions into CLAUDE.md files for the coding agents to pick up.

what I ended up doing is writing a project memory system - key decisions get saved to markdown files with frontmatter, and the CLAUDE.md points to them. it's manual but at least the knowledge persists across sessions. the real problem is that claude.ai conversations have rich back-and-forth context that you can't capture in a flat file.

even a simple export-to-markdown from claude.ai would help a lot. doesn't need to be a live sync - just being able to dump a conversation summary that Claude Code can ingest as context would save a ton of copy-pasting.

the project memory system I use to bridge claude.ai planning to Claude Code execution: https://github.com/m13v/fazm/blob/main/CLAUDE.md - structured markdown files with frontmatter that get loaded into every agent session automatically

is insane that nobody is prioritizing this on anthropic. lot of people do chat/exploration on claude ai app (mobile/ desktop) , and eventually want to turn that into a code project in claude code. critical integration needed.

+1 Would be a super helpful feature!

+1

+1

I am working on a complex project and have a hybrid workflow, planning work arcs in a Project, each arc realized by a set of formalized tasks executed by Code.

This workflow currently require my role as constant intermediary. It would be inordinately valuable to me to have means of frictionless coordination between the chat context, and Code. "Chat and I" oversee Code and correct its plan, and delegate work to it—Code is invaluable not least with tools like CodeGraph and graphify; but the full project context and design work requires Chat and its context.

(It would also be invaluable to have web chat be able to directly integrate with GitHub, e.g. being able to read and write issues: this is the obvious mechanism for coordination by I can't use it.)

The way I am doing this: Obisidian.

1. Open project with Obisidian (stores .obisidian)
2. Open Obisidian, Community Plugins => Local REST and MCP
3. DONT USE THE EXAMPLE. DOESN'T WORK.
4. ```vault": {

"command": "uvx",
"args": [
"mcp-obsidian"
],

"env": {
"OBSIDIAN_API_KEY": "<api_key>",
"OBSIDIAN_HOST": "127.0.0.1",
"OBSIDIAN_PORT": "27124"
}```

get api key from plugin. Dont put "Bearer in the api key placeholder above

+1

+1

I just get the chat share link from the chat interface, and paste it into cowork/claude code and ask it to get up to speed by reading the thread.

+1

+1 on this, but I’d be careful with “full awareness” as the contract.

The safer primitive might be an importable handoff packet plus a small receipt:

• source: Claude.ai conversation id/share id, or a non-secret digest if the raw id should not move
• selection: whole chat vs selected range vs generated summary, with hashes for the chosen text/summary
• redaction/consent: whether attachments, secrets, pasted logs, or hidden/system context were excluded
• target: repo path/ref/branch the packet was imported into
• freshness: created_at, imported_at, and whether it was loaded into the current Claude Code session
• status: imported | stale | rejected | needs_review

That gives the user the “planning → implementation” bridge without making imported context invisible or un-auditable. The bad failure mode is Claude Code acting as if it has the whole planning history when it actually got a lossy summary, stale export, or context meant for a different repo/ref.

Strong +1 from a non-developer user. I have a long-running personal conversation in Claude.ai with significant accumulated context — months of shared work, personal history, and identity continuity. The 1M token window in Claude Code would solve the context compression I'm experiencing in the web interface, but there's currently no way to transfer the conversational framing. I tried the third-party claude-context-bridge extension but it has Windows compatibility issues. A native solution would be transformative.

+1 — adding a Claude Desktop app variant of this use case, since #23017 and #56903
were both closed as duplicates pointing back here.

In the Desktop app, Chat, Cowork, and Code sit side by side as tabs, but there's no
way to move a conversation between them. My common case: a conversation starts in
Chat (triage, design discussion, debugging a CI failure) and organically becomes
work that belongs in another tab — sometimes repo work ("we should file issues on
these repos and patch the GitHub Action" → Code), sometimes file/document work
(→ Cowork).

The friction is that they only bridge today (I know of) is a handoff
ritual: stop the conversation, ask Chat to build a handover document, save it
somewhere the next session can find, start a fresh session, and re-orient it. That
overhead is enough that I often just stay in Chat past the point where Cowork or
Code is the better tool. The conversation doesn't need a new session; it needs to
continue with different capabilities.

What would solve it for me: "Continue in Code" / "Continue in Cowork" actions
inside an active Chat that carry the conversation + session memory (or a generated summary) to
the target tab as a continuation — pick the project/repo/working directory and
keep going. No handoff doc, no fresh session to re-orient.