Share, Embed, and Curate Agent Sessions on DEV [Beta]
![Share, Embed, and Curate Agent Sessions on DEV [Beta]](https://media2.dev.to/dynamic/image/width=1000,height=420,fit=cover,gravity=auto,format=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2Fwk6kbgjbme42vp7aotnx.png)
Authors can now upload Agent Session transcripts to their DEV account. Those transcripts can be curated, sliced up, shared publicly, and embedded into posts. This is a beta feature, and we are actively looking for feedback!
Every day, more people are creating software for the first time. Some have long histories as developers, and some have never seen a line of code in their life. In most cases, they're using the same tooling - coding agents. And yet, as these tools become ubiquitous there are surprisingly few ways to share your workflows and results in a standardized way beyond screenshots. Each tool has its own share link generator, each tool outputs logs in a different format. There are a few nascent projects out there to help, but they generally exist as standalone utilities or don't solve the sharing problem explicitly in a way that works on DEV.
We decided to solve that problem by making it possible to upload your transcripts to DEV, automatically parse them into a standardized format, and embed whole or partial sessions into your posts in a way that fits naturally into technical writing.
Supported Platforms
To start with, our parser supports Gemini CLI, Codex, Claude Code, Pi, and GitHub Copilot CLI. If you use a different tool, you're encouraged to open a PR or Issue on Forem to add a new parser yourself or ask us to do so. As an example, you can check out how our Gemini CLI parser works in our GitHub commit:
https://github.com/forem/forem/commit/8f92c64a775ed3a0993e5edc0637112a95d1dae2#diff-d98a98bdd6e0d9154df5559b53644185d4ab13160e85c0fdaae3013ff985e27b
Example Session
Here's a session from an example app I built using Gemini CLI for my pet guinea pig Pearl (obligatory pic in comments).
build me a simple web app to let my guinea pig tell me what food it wants
Clarifying Application Needs / Defining Food Selection UI / Mapping Out The App's Flow
I'll build a vibrant, touch-friendly web app called "Pui Pui Picker" designed for guinea pigs (and their humans).
The Plan:
- Scaffold: Create a React + TypeScript project using Vite.
- Design: A "big button" interface with high-contrast colors and bouncy animations to make it engaging.
- Features:
- Food Grid: Large cards for Hay, Carrots, Romaine, and Bell Peppers.
- Pui Pui Feedback: Visual and text-based "wheeking" feedback when an item is selected.
- History: A small "Last Requested" log to track what the piggie "chose."
- Styling: Custom Vanilla CSS for a polished, modern look without external dependencies.
Does this sound like a good start for your guinea pig? Or should I add specific snacks they love?
I will start by scaffolding the project.
{
"command": "npm create vite@latest . -- --template react-ts && npm install",
"description": "Scaffolding the React TypeScript project using Vite and installing dependencies."
}let's go simpler than vite and react
Seeking Simplification for App / Constructing the index.html
I'll create a single-file, vibrant, and touch-friendly HTML application called "The Wheeky Menu". It will use large, colorful buttons and bouncy animations—perfect for a guinea pig's "input" (or a human's finger).
The Plan:
- Single File: Everything (HTML, CSS, JS) in
index.htmlfor instant loading. - Piggie-Centric UI:
- Large Interaction Zones: Easy-to-hit cards for Hay, Carrots, Peppers, and Lettuce.
- Visual Feedback: Cards will "pop" and flash when tapped.
- Features:
- "Wheek!" Notification: A large visual alert showing what was picked.
- Order History: A simple log at the bottom to see what the piggie "requested" most.
I will now create the index.html file.
{"content":"\u003c!DOCTYPE html\u003e\n\u003chtml lang=\"en\"\u003e\n\u003chead\u003e\n \u003cmeta charset=\"UTF-8\"\u003e\n \u003cmeta name=\"viewport\" content=\"width=device-width, initial...Successfully created and wrote to new file: /Users/jonmarkgo/Documents/GitHub/gemini-guinea-pig/index.html. Here is the updated code:
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0, maximum-scale=1.0, user-scalable=no">
<title>Wheeky Menu - Piggie Food Picker</title>
<style>
:root {
--bg-color: #fdf6e3;
--card-bg: #ffffff;
--text-color: #5d4037;
--accent-color: #ff8a65;
--shadow: 0 8px 16px rgba(0,0,0,0.1);
--color-hay: #fff176;
--color-carrot: #ffb74d;
--color-pepper: #ef5350;
--color-lettuce: #81c784;
}
* {
box-sizing: border-box;
-webkit-tap-highlight-color: transparent;
}
body {
font-family: 'Segoe UI', Tahoma, Geneva, Verdana, sans-serif;
background-color: var(--bg-color);
color: var(--text-color);
margin: 0;
display: flex;
flex-direction: column;
align-items: center;
min-height: 100vh;
padding: 20px;
overflow-x: hidden;
}
header {
text-align: center;
margin-bottom: 30px;
}
h1 {
font-size: 2.5rem;
margin: 0;
color: var(--accent-color);
text-shadow: 2px 2px 0px #fff;
}
p.subtitle {
font-style: italic;
margin-top: 5px;
opacity: 0.8;
}
.grid {
display: grid;
grid-template-columns: repeat(2, 1fr);
gap: 20px;
width: 100%;
max-width: 600px;
}
.food-card {
background: var(--card-bg);
border-radius: 24px;
padding: 30px 10px;
text-align: center;
box-shadow: var(-
... (truncated)Constructing The Interface / Refining The UX Design
The "Wheeky Menu" is ready! It's a single-file web app with large, high-contrast buttons designed for easy "selection" (by snout or finger).
How to use:
- Open the file: You can open
index.htmldirectly in any web browser. - Piggie Input: Place your tablet or phone on the floor or near the cage. When your guinea pig taps or "boops" a card:
- A large "WHEEK!" announcement pops up.
- The request is added to the Request History at the bottom (so you can see what they picked if you step away).
- On mobile devices, it will provide a small vibration (haptic feedback) when touched.
I've used high-contrast colors and bouncy animations to make it visually stimulating. Enjoy the wheeking!
Find the current Gemini CLI session log file. Look in ~/.gemini/sessions/ for the most recently modified .json or .jsonl file.
Then run the devto command to upload it:
Error: No file specified. Use --auto or provide a file path.\nRun 'devto --help' for usage.
[Shell command 'devto --yes --tool gemini_cli --title "" "$(ls -t ~/.gemini/sessions//*.json ~/.gemini/sessions//*.jsonl 2>/dev/null | head -1)"' exited with code 1]
If devto is not installed, use curl directly:
jq: Bad JSON in --rawfile body : Could not open : No such file or directory
[Shell command 'TMPJSON=$(mktemp) && jq -n --arg title "" --rawfile body "$(ls -t ~/.gemini/sessions//*.json ~/.gemini/sessions//*.jsonl 2>/dev/null | head -1)" '{title: $title, tool_name: "gemini_cli", body: $body}' > "$TMPJSON" && curl -s -X POST "${DEVTO_HOST:-https://dev.to}/api/agent_sessions" -H "Content-Type: application/json" -H "Accept: application/vnd.forem.api-v1+json" -H "api-key: $DEVTO_API_KEY" -d @"$TMPJSON" && rm -f "$TMPJSON"' exited with code 2]
After uploading, display the URL of the created session to the user.
Examining Upload Strategies / Implementing Dev Upload Skill
I will activate the dev-upload skill to securely and efficiently upload your session log to DEV.to.
{
"name": "dev-upload"
}<activated_skill name="dev-upload">
<instructions>
# Upload Session to DEV
Upload the current session transcript to [DEV](https://dev.to) as an Agent Session.
## Steps
### 1. Find the transcript
Find the current session's transcript file — the most recently modified `.jsonl` file:
- Claude Code: `~/.claude/projects/` (recursive)
- Codex: `~/.codex/sessions/`
- Gemini CLI: `~/.gemini/sessions/`
- GitHub Copilot: `~/.copilot/`
- Pi: `~/.pi/sessions/`
Use the agent's own session context if available to identify the exact path.
### 2. Load config and check for API key
Load all config in a single bash call:
```bash
if [ -f ~/.config/devto/config ]; then source ~/.config/devto/config; fi
API_KEY="${DEVTO_API_KEY:-${API_KEY:-}}"
HOST="${DEVTO_HOST:-https://dev.to}"
echo "API_KEY=${API_KEY:+SET}"
echo "HOST=$HOST"
```
If no key is found:
1. Tell the user they need a DEV API key and link them to https://dev.to/settings/extensions to generate one.
2. Ask the user to paste their key here in the chat.
3. Once they provide it, save it (along with the host) so they never have to do this again:
```bash
mkdir -p ~/.config/devto
cat > ~/.config/devto/config <<'CONF'
DEVTO_API_KEY="THE_KEY_THEY_GAVE_YOU"
DEVTO_HOST="https://dev.to"
CONF
```
4. Then continue with the upload. Do NOT stop or ask the user to run commands themselves.
### 3. Upload
Do the entire upload in a single bash call. Replace TRANSCRIPT, TITLE, HOST, and API_KEY with actual values:
```bash
set -e
TRANSCRIPT="/path/to/session.jsonl"
TITLE="the session title"
HOST="https://dev.to"
API_KEY="the-key"
TMPJSON=$(mktemp)
trap "rm -f '$TMPJSON'" EXIT
jq -n \
--arg title "$TITLE" \
--arg tool_name "auto" \
--rawfile body "$TRANSCRIPT" \
'{title: $title, tool_name: $tool_name, body: $body}' > "$TMPJSON"
RESPONSE=$(curl -s -w "\n%{http_code}" \
-X POST "$HOST/api/agent_sessions" \
-H "Content-Type: application/json" \
-H "Accept: application/vnd.forem.api-v1+json" \
-H "api-key: $API_KEY"
... (truncated)Defining the Upload Process / Analyzing the Configuration Steps
I'll locate your most recent session log and check for a DEV.to API key to begin the upload.
{"description":"Locating the most recent session log and checking for a DEV.to API key.","command":"# Find the most recent session log\nTRANSCRIPT=$(ls -t ~/.gemini/sessions/*.jsonl 2\u003e/dev/nul...Output: TRANSCRIPT=
API_KEY=SET
HOST=https://dev.to
Process Group PGID: 73065Searching the File System / Refining the Search Strategy
I'll look for the session log again, including both .json and .jsonl files across all session subdirectories.
{"command":"# Find the most recent session log\nTRANSCRIPT=$(ls -t ~/.gemini/sessions/**/*.json ~/.gemini/sessions/**/*.jsonl 2\u003e/dev/null | head -1)\necho \"TRANSCRIPT=$TRANSCRIPT\"","descript...Output: TRANSCRIPT=
Process Group PGID: 73508Exploring the Directory
I'll list the contents of ~/.gemini/sessions/ to find the exact session log files.
{
"command": "ls -R ~/.gemini/sessions/ | head -n 20",
"description": "Listing the contents of the Gemini CLI sessions directory."
}Output: ls: /Users/jonmarkgo/.gemini/sessions/: No such file or directory
Process Group PGID: 73515Examining Session Locations / Inspecting Gemini Directory Contents
I'll list the contents of the ~/.gemini/ directory to locate the session logs.
{
"command": "ls -F ~/.gemini/",
"description": "Listing the contents of the ~/.gemini directory."
}Output: antigravity/ google_accounts.json mcp-oauth-tokens-v2.json settings.json tmp/
commands/ history/ oauth_creds.json settings.json.orig trustedFolders.json
GEMINI.md installation_id projects.json state.json user_id
Process Group PGID: 73530Checking the History
I'll list the contents of ~/.gemini/history/ to find the session log.
{
"command": "ls -F ~/.gemini/history/ | head -n 20",
"description": "Listing the contents of the ~/.gemini/history directory."
}Output: cookieparser/
gemini-dog-app/
gemini-guinea-pig/
mlh-core/
Process Group PGID: 73565Examining Project Location
I'll list the contents of ~/.gemini/history/gemini-guinea-pig/ to find the current session's log.
{
"command": "ls -t ~/.gemini/history/gemini-guinea-pig/*.jsonl 2>/dev/null | head -1",
"description": "Searching for the most recent session log in the history directory."
}Output: (empty)
Process Group PGID: 74637Beginning File Enumeration
I'll list all files in ~/.gemini/history/gemini-guinea-pig/ to find the session log.
{
"command": "ls -R ~/.gemini/history/gemini-guinea-pig/ | head -n 100",
"description": "Listing all files in the history folder to locate the log file."
}Output: (empty)
Process Group PGID: 75100Pinpointing the Missing Path / Re-evaluating the Scope
I'll list the contents of ~/.gemini/tmp/gemini-guinea-pig/ to find the session log.
{
"command": "ls -F ~/.gemini/tmp/gemini-guinea-pig/ | head -n 20",
"description": "Listing the contents of the ~/.gemini/tmp/gemini-guinea-pig directory."
}Output: chats/
logs.json
Process Group PGID: 75368Investigating Log Location
I'll read the first 10 lines of logs.json to see if it's the session log.
{
"command": "head -n 10 ~/.gemini/tmp/gemini-guinea-pig/logs.json",
"description": "Reading the beginning of logs.json to confirm its format."
}Output: [
{
"sessionId": "295d2fe2-6276-49c4-9229-db8a61d33f2a",
"messageId": 0,
"type": "user",
"message": "build me a simple web app to let my guinea pig tell me what food it wants",
"timestamp": "2026-03-04T01:15:24.210Z"
},
{
"sessionId": "295d2fe2-6276-49c4-9229-db8a61d33f2a",
Process Group PGID: 75447Investigating the Discrepancy / Unraveling File Paths
I'll look for any other directories starting with 's' in ~/.gemini/ to find the sessions folder.
{
"command": "ls -d ~/.gemini/s*",
"description": "Listing directories in ~/.gemini that start with 's'."
}Output: /Users/jonmarkgo/.gemini/settings.json /Users/jonmarkgo/.gemini/settings.json.orig /Users/jonmarkgo/.gemini/state.json
Process Group PGID: 75492Examining Possible Locations
I'll list all directories starting with .gemini in my home directory to find the session logs.
{
"command": "ls -d ~/.gemini*",
"description": "Listing all directories and files that start with .gemini in the home folder."
}Output: /Users/jonmarkgo/.gemini
Process Group PGID: 75500Searching for Session Logs / Identifying Alternative Locations
I'll search for all .json and .jsonl files in ~/.gemini/ to find the correct session log.
{
"description": "Searching recursively for .json and .jsonl files in the ~/.gemini directory.",
"command": "find ~/.gemini/ -name \"*.json\" -o -name \"*.jsonl\" | head -n 20"
}Output: /Users/jonmarkgo/.gemini/google_accounts.json
/Users/jonmarkgo/.gemini/state.json
/Users/jonmarkgo/.gemini/oauth_creds.json
/Users/jonmarkgo/.gemini/settings.json
/Users/jonmarkgo/.gemini/projects.json
/Users/jonmarkgo/.gemini/trustedFolders.json
/Users/jonmarkgo/.gemini/antigravity/mcp_config.json
/Users/jonmarkgo/.gemini/mcp-oauth-tokens-v2.json
/Users/jonmarkgo/.gemini/tmp/08d27515e40a8c790d53798895cf6650d5843987a3978f7ef5be01987a09b3e1/logs.json
/Users/jonmarkgo/.gemini/tmp/a9c39dae5288a24cdef1cf1d8c6bc99a5cf8b751707de794545fc5f5986be7b2/logs.json
/Users/jonmarkgo/.gemini/tmp/dd9dc36095d552fc5d0c79d94e889e2c50c1b76a76ea3f41ac6b22dd6076ce19/logs.json
/Users/jonmarkgo/.gemini/tmp/dd9dc36095d552fc5d0c79d94e889e2c50c1b76a76ea3f41ac6b22dd6076ce19/chats/session-2025-12-04T21-23-e153598e.json
/Users/jonmarkgo/.gemini/tmp/gemini-guinea-pig/logs.json
/Users/jonmarkgo/.gemini/tmp/gemini-guinea-pig/chats/session-2026-03-04T01-15-295d2fe2.json
/Users/jonmarkgo/.gemini/tmp/gemini-guinea-pig/chats/session-2026-03-04T12-42-295d2fe2.json
/Users/jonmarkgo/.gemini/tmp/79261791c475d806f2ff6cdf73b50843055032e35b2854d7e91e3a92762f2e96/logs.json
/Users/jonmarkgo/.gemini/tmp/79261791c475d806f2ff6cdf73b50843055032e35b2854d7e91e3a92762f2e96/chats/session-2025-12-06T18-02-b9fe7181.json
/Users/jonmarkgo/.gemini/tmp/79261791c475d806f2ff6cdf73b50843055032e35b2854d7e91e3a92762f2e96/chats/session-2025-11-18T16-06-1f38a3eb.json
/Users/jonmarkgo/.gemini/tmp/79261791c475d806f2ff6cdf73b50843055032e35b2854d7e91e3a92762f2e96/chats/session-2025-11-18T18-32-22205976.json
/Users/jonmarkgo/.gemini/tmp/79261791c475d806f2ff6cdf73b50843055032e35b2854d7e91e3a92762f2e96/chats/session-2025-11-18T16-52-0863224c.json
Process Group PGID: 76229Analyzing File Locations / Pinpointing Key Files
I'll select the latest session log from ~/.gemini/tmp/gemini-guinea-pig/chats/ and upload it to DEV.to using the discovered session file and your API key.
Let's check out how the upload, curation, and embedding workflow works.
Upload a Session
First, you need to upload a session. You can drag and drop your session transcripts via your browser on the Agent Sessions page. We will soon be launching a CLI and Agent skill to give you a slash command for uploads as well.
Curate a Session
Once you upload the session, you can curate it. You can select which pieces of a session transcript are included in your whole-session embed. You can mass enable/disable different pieces of your transcript, as shown here.
Slice Up a Session
If you don't want to embed your whole session (even in curated form) you can instead use the Slice tool to create a sub-transcript that can include any and all parts of a session. You can have multiple Slices per session, if you want to break up your prompt and embed it in different places in your post.
Share a Session
Lastly, to make this utility more generally useful beyond DEV, you can also share curated Agent Sessions publicly by using the Make Public button. See my session here.
Note: Please be very cognizant of what potentially sensitive data/keys might be in your transcripts. We have a redaction layer built in to the parser, but it is not going to catch every edge case. Do not upload sessions or data that you do not own, or that may open you up to security vulnerabilities.
We look forward to seeing what you build and share with this new feature. Happy hacking!