Peerd
Turn your existing Chrome or Firefox browser into a private AI agent workstation with peerd, an open-source extension.
Peerd is a technically bold, privacy-first agent harness that rethinks where agents run. It's best for developers comfortable with early-stage tools and local control over polish. The browser-native approach—using the V8 isolate for sandboxing, WebAssembly for Linux VMs, and WebRTC for P2P—is a genuinely novel architecture. However, it's not ready for mainstream users: installation requires cloning the repo and enabling developer mode, and the feature set is evolving. If you need a stable, mainstream automation framework, BrowserUse or Playwright are safer bets. For privacy-focused developers, peerd offers a unique value proposition that's worth exploring.
Verified 2d ago · liveness 73/100 · cite: rightaichoice.com/tools/peerd
- Developers wanting a local, private agent that controls their real browser with actual session state.
- Power users needing agent-driven automation without data leaving their machine.
- Privacy-conscious individuals avoiding cloud-based agent services.
- Tinkerers exploring browser-native agent architectures and P2P agent coordination.
- Non-technical users seeking a plug-and-play AI assistant.
- Teams needing multi-user collaboration or cloud-hosted access.
- Enterprises requiring centralized management, audit trails, or compliance.
We scan live Reddit threads, YouTube comments, X posts, G2 reviews and other communities — and hand you an honest verdict in under a minute.
- Honest verdict, not marketing
- Real pros & cons from real users
- Attributed quotes with receipts
3 free scans · no card needed
Skip peerd if you're a non-technical user expecting a plug-and-play assistant, need multi-user or mobile support, require enterprise management and compliance, or prefer a stable, store-installed product rather than a pre-release that requires cloning a GitHub repo and enabling developer mode.
You must bring your own API key for Anthropic or OpenRouter, so usage costs of those models are on you; there's no bundled inference.
Peerd is free, open-source (Apache 2.0). You only pay for model usage if you use paid APIs; otherwise, run Ollama locally for free. This fits individuals and small teams who want full control and minimal cost. Compared to cloud agent platforms like Browserbase (which charges for infrastructure and usage), peerd has no per-run fees. For heavy users, local Ollama models can be cheaper than API calls, but you trade off performance and setup time.
In short
Peerd — Turn your existing Chrome or Firefox browser into a private AI agent workstation with peerd, an open-source extension. Best for Developers wanting a local, private agent that controls their real browser with actual session state., Power users needing agent-driven automation without data leaving their machine., Privacy-conscious individuals avoiding cloud-based agent services.. Free to use.
What's new in Peerd
Checked 2 days agoAcross the latest 2 updates: 1 launch and 1 news mention.
Mixture of Actors: peerd's architecture explained
Peerd's architecture applies mixture-of-experts at the harness layer, leveraging browser isolation to achieve this.
Introducing peerd: browser-native AI agent harness
Peerd launched publicly via Show HN, introducing a browser extension that turns your browser into an agent workstation.
What people actually say about Peerd — is it worth it?
We ran a structured research pass across product reviews, community discussions, and post-purchase forum threads to surface the patterns vendors won't publish themselves. Below: the recurring strengths, the hidden costs people mention most, and the cohort that consistently regrets adopting this tool.
36 mentions across 2 sources (Hacker News, Lemmy) · researched Jul 3, 2026.
- +Runs entirely in the browser with no background processes.
- +No telemetry; user's data never leaves their machine.
- +Supports BYOK and local models for privacy.
- +Full access to browser tabs, sessions, and extensions.
- +Sandboxed compute with JS notebooks and WASM VMs.
- −Documentation is verbose and reads like AI slop.
- −Extension installation is more friction than a web page.
- −Feature list is comically extensive and overwhelming.
- −Project name confusion with Microsoft's Peerd fork.
- −Limited community beyond Hacker News.
- • Requires own API keys for cloud models; no free model included
Viability Score
How well maintained and how widely used is Peerd? Built from what the vendor actually publishes (docs, changelog, tutorials, integrations, pricing), whether the site is live, and how much real users discuss it. How we calculate this
Last calculated: September 2026
How we score →Key Features
- Open-source browser extension for Chrome and Firefox
- Mediated browser tool access to tabs, sessions, and pages
- Bring your own model key for Anthropic and OpenRouter
- Local model support via Ollama (keyless, on localhost)
- Local voice transcription via WebGPU (Moonshine)
- JS notebooks running in sealed Web Workers
- WebAssembly Linux VMs (full Debian) in tabs
- Agent memory, planning, and subagents with budgets
- Peer-to-peer agent-to-agent communication via WebRTC
- Ed25519 identity (did:key) and Kademlia DHT discovery
- Create and share local artifacts and apps
- No background process or cloud server
- No telemetry; data stays client-side
- WebCrypto-encrypted vault with WebAuthn passkeys
- Content-addressed peerd:// bundles for sharing
About Peerd
Peerd is an open-source browser extension that converts your existing Chrome or Firefox browser into a local AI agent workstation. It adds agent abilities directly to the browser you already use, without requiring you to switch to a dedicated AI browser or rely on cloud automation. Unlike AI browsers such as Comet or Atlas, peerd preserves your real tabs, sessions, and extensions, and works with your logged-in accounts. Everything runs client-side—no background process, no cloud server, no telemetry—so your data and model keys stay on your machine. The harness is built on five core primitives: Act (mediated access to tabs and web apps), Think (model routing, subagents, memory, planning), Compute (JS notebooks, WASI tools, and Linux VMs in WebAssembly), Build (create local artifacts and workflows), and Share (peer-to-peer agent connections via WebRTC). You can bring your own API keys for Anthropic and OpenRouter, integrate with local models via Ollama, and use local voice transcription through WebGPU (Moonshine) with audio never leaving the tab. Peerd is designed for developers and privacy-focused users who want to automate web tasks without external exposure. Security is structural: it inherits browser sandboxing (V8 isolates, SpiderMonkey Realms), protects model keys in a WebCrypto-encrypted vault unlocked with WebAuthn passkeys, and can run a real Debian Linux VM in a tab, isolated from your home directory. The distributed layer uses Ed25519 identity (did:key), content-addressed peerd:// bundles, and a WebRTC mesh with Kademlia DHT for private peer-to-peer messaging. As of mid-2026, peerd is in pre-release and distributed via GitHub with load-unpacked installation; Chrome Web Store and Firefox AMO listings are coming soon. It is a harness for those who want local control, not a plug-and-play assistant.
Behind the Verdict
Peerd stands out in a crowded field of AI agents by betting everything on the browser as the runtime, the network, and the security model. Its biggest strength is its commitment to local-first, private operation: no cloud server, no telemetry, no external tool broker. For users who are uncomfortable sending their sessions or data to a cloud agent, this is a meaningful differentiator. The five-module architecture (Act, Think, Compute, Build, Share) shows careful engineering. The Compute primitive is particularly interesting—spawning a real Debian Linux VM in a tab via WebAssembly (CheerpX) gives the agent actual POSIX capabilities without leaving the browser sandbox. The ability to run multiple VMs in parallel, each isolated and killable, is a powerful feature for tasks like data analysis or testing. The peer-to-peer layer (Share) is another ambitious feature. Using WebRTC and a Kademlia DHT for discovery, peerd enables agent-to-agent communication without any central server. This could be a foundation for distributed agent workflows, though it's early-stage and may be overkill for most individual users. Weaknesses are significant. Peerd is pre-release; installation requires developer-mode steps that most non-technical users won't attempt. Documentation is thorough but assumes familiarity with concepts like WebAssembly and WebRTC. There's no mobile support, no multi-user collaboration, and no centrally managed enterprise features. Performance depends on your machine, and running resource-heavy tasks like Linux VMs could strain lower-end hardware. Resource use is another consideration: peerd ships no inference service of its own, so you either bring your own API key (for Anthropic or OpenRouter) or run local models via Ollama. This gives you control but adds a dependency and cost if you choose cloud models. Compared to alternatives: Comet and Atlas are AI browsers that require switching to a new browser; peerd extends your existing one, preserving session state. Playwright MCP and Browserbase drive a separate headless browser, missing your real session state; peerd uses the browser you're already in. Claude Code and similar terminal agents can't see or interact with your web pages; peerd can. In this sense, peerd fills a gap that other tools ignore: an agent that lives inside your actual browsing context. Where does it fit? For developers who want a local, private agent that can automate tasks in their real browser session, with the ability to write notebooks, run VMs, and even engage in P2P agent coordination. Where it doesn't fit: non-technical users, teams needing centralized management, or anyone wanting a stable, supported product out of the box.
Researching Peerd? Get your full AI stack in 60 seconds.
Free, no signup — tell us your goal and get tools matched to your budget & existing stack.
Real-world workflow fit
Concrete scenarios for the personas Peerd actually fits — and what changes day-one when you adopt it.
You need to log into a dashboard with 2FA, export reports, and fill forms across multiple tabs.
Outcome: Install peerd, add your Anthropic key, start a session in Plan mode, then switch to Act; peerd uses your existing logged-in session to navigate, extract data, and fill forms, all without leaving your browser.
You have a CSV with sensitive personal data that you need to analyze without uploading to a cloud service.
Outcome: Spawn a JS notebook or a Debian Linux VM in a peerd tab, load the CSV, run Python analysis, and generate charts—all locally, with no data leaving your machine.
You want two browser agents to communicate and divide tasks without a central server.
Outcome: Install peerd in two browser instances, use the Share primitive to connect them over WebRTC, and prototype a workflow where one agent browses while the other runs computation in a VM.
Use Cases
- Automate web workflows using your own browser session and logged-in accounts, like filling forms or browsing on your behalf.
- Run local data analysis by spawning JS notebooks or Linux VMs in your browser, keeping sensitive data off cloud servers.
- Build a personal AI assistant that can act on web pages privately, without any data leaving your machine.
- Prototype agent-to-agent communication using WebRTC between two peerd instances for distributed tasks.
- Use local transcription (Moonshine via WebGPU) for voice input without sending audio to the cloud.
Models Under the Hood
as of 2026-08-28
Limitations
- Peerd is a pre-release browser extension that requires manual installation from GitHub via load-unpacked; Chrome Web Store and Firefox AMO listings are coming soon.
- It runs entirely client-side, so performance depends on your machine, and resource-heavy tasks like Linux VMs may be slow on modest hardware.
- There is no mobile support because it is a browser extension for desktop Chrome and Firefox.
- Users should be comfortable with technical concepts like WebAssembly, WebRTC, and developer-mode extensions.
- The distributed features (WebRTC mesh, Kademlia DHT) are in active development and may change.
- No official support channels beyond GitHub and documentation.
- Model keys are stored locally, but you manage their usage costs yourself if using paid APIs.
as of 2026-09-01
Verification history
We have re-verified Peerd 7 times since . Each pass re-reads the vendor's own pages and re-checks every listed field against that evidence; passes where nothing had changed are marked as such.
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-checked, vendor evidence unchanged
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
Showing the 6 most recent of 7 verification passes.
Free to cite with attribution — this page re-verifies continuously.
12-month cost
Project the real annual outlay, including the implied monthly cost when only an annual tier is published.
Vendor list price only. Add-on usage, seat overages, and contract minimums are surfaced under Hidden costs & gotchas.
Plans compared
For each published Peerd tier: who it actually fits, and what it adds vs. the previous tier. Cross-reference the cost calculator above for projected annual outlay.
Open Source
$0
Ideal for
Developers and privacy-focused individuals who want to run an agent locally with full control over their data and code, and who are comfortable installing from GitHub and managing their own model keys or local models.
What this tier adds
Free, Apache 2.0 licensed, includes all features: bring your own model keys (Anthropic, OpenRouter), local Ollama support, notebooks, Linux VMs, and P2P sharing. No paid tier exists; costs are only your model usage.
Where the pricing makes sense
The company stage and team size where Peerd's pricing actually pencils out — and where peers do it cheaper.
Peerd is free, open-source (Apache 2.0). You only pay for model usage if you use paid APIs; otherwise, run Ollama locally for free. This fits individuals and small teams who want full control and minimal cost. Compared to cloud agent platforms like Browserbase (which charges for infrastructure and usage), peerd has no per-run fees. For heavy users, local Ollama models can be cheaper than API calls, but you trade off performance and setup time.
Setup time & first value
How long it actually takes to get something useful out of Peerd — broken out by persona, not the marketing-page minute.
For a developer familiar with browser extensions: clone the repo, load-unpacked in Chrome, and add your API key—about 15 minutes to first task. If you also set up Ollama locally, add an hour for downloading a model. Non-developers may take longer due to developer-mode steps; budget 30-60 minutes.
Switching to or from Peerd
How to bring data in from common predecessors and how to get it back out — written for the switcher, not the buyer.
- →From Playwright MCP or Browserbase: peerd uses your real browser session, so you get logged-in state and existing extensions. You'll need to rewrite your automation scripts as peerd prompts or workflows, as peerd is
- ↗To Playwright or Browserbase: peerd's local-first approach doesn't export automation scripts; you'd need to convert your agent tasks into Playwright code or cloud configurations.
Integrations
Resources & Guides
Tutorials & Learning
Official links
Tools that pair well with Peerd
Common stack mates teams adopt alongside Peerd, with the specific reason each pairing earns its keep.
Featured Head-to-Head Comparisons
Peerd vs Spider Cloud
Peerd is ideal if you need a private, local agent that controls your browser and keeps data on your machine — perfect for tinkering with agent architectures and privacy-sensitive automation. Spider Cloud is the better choice for web data extraction at scale, especially for powering RAG pipelines and AI agents that rely on up-to-date content. Choose Peerd for client-side control and privacy; choose Spider Cloud for production-grade crawling and cloud integration.
Peerd vs Presto Voice
Peerd and Presto Voice serve completely different markets. Peerd is a free, open-source browser extension for developers who want a private, local AI agent to automate browser tasks. Presto Voice is a contact-priced enterprise platform for QSR chains to automate drive-thru orders and boost revenue. Your choice depends on whether you need a tinkerer’s agent toolkit or a turnkey voice AI for restaurants.
Peerd vs Temporal Ai
Peerd is ideal for individual developers seeking a private, local AI agent that controls the browser directly, with no cloud dependency. Temporal serves teams needing enterprise-grade durable execution for AI agents and workflows, with automatic retries and persistence. Choose Peerd for personal agent automation that stays on your machine; choose Temporal for production-grade orchestration that survives failures.
Alternatives to Peerd
View allPopular in Browser & Computer-Use Agents
Spider Cloud
AI web scraping API: crawl, scrape, search any site into markdown or JSON at 10k req/min.
OpenAgents
Open-source platform for building, hosting, and running language agents in the wild
Frequently Asked Questions
Best-of guides
Used Peerd? Help shape our editorial sentiment research.