{"_id":"@antonlx/perplexity-web-api-mcp","_rev":"2-5f06e14564123cd4d53bed9034b00f39","name":"@antonlx/perplexity-web-api-mcp","dist-tags":{"latest":"0.12.2"},"versions":{"0.12.1":{"name":"@antonlx/perplexity-web-api-mcp","version":"0.12.1","keywords":["perplexity","mcp","ai","search"],"license":"MIT","_id":"@antonlx/perplexity-web-api-mcp@0.12.1","maintainers":[{"name":"antonlx","email":"antonix@devpins.org"}],"homepage":"https://github.com/AntonIXO/perplexity-web-api-mcp#readme","bugs":{"url":"https://github.com/AntonIXO/perplexity-web-api-mcp/issues"},"bin":{"perplexity-web-api-mcp":"run-perplexity-web-api-mcp.js"},"dist":{"shasum":"22cbdc93d840394eb0893d0285e5991352ad54eb","tarball":"https://registry.npmjs.org/@antonlx/perplexity-web-api-mcp/-/perplexity-web-api-mcp-0.12.1.tgz","fileCount":11,"integrity":"sha512-JaVi8I46bU6aE/Zn8IQ5OYcvyOmCnNHjrK0Pj08SiXAa3ykkK6jnsrnWTDRfUNrFHhJLTV7sNpQ7ZiSRHiH/sQ==","signatures":[{"sig":"MEUCIQCIIfG4GwNzmQb+sXomYg2C6jS31RY6vJY4qYvei4XJ0QIgaWDwsdvORJ25oWVeNWR5MFrqXQ1ho1hNlTEKX+ad4Jo=","keyid":"SHA256:DhQ8wR5APBvFHLF/+Tc+AYvPOdTpcIDqOhxsBHRwC7U"}],"unpackedSize":55576},"_from":"file:npm/perplexity-web-api-mcp-npm-package.tar.gz","volta":{"npm":"9.5.0","node":"18.14.1"},"engines":{"npm":">=6","node":">=14"},"scripts":{"fmt":"prettier --write **/*.js","fmt:check":"prettier --check **/*.js","postinstall":"node ./install.js"},"_npmUser":{"name":"antonlx","email":"antonix@devpins.org"},"_resolved":"/home/runner/work/perplexity-web-api-mcp/perplexity-web-api-mcp/npm/perplexity-web-api-mcp-npm-package.tar.gz","_integrity":"sha512-JaVi8I46bU6aE/Zn8IQ5OYcvyOmCnNHjrK0Pj08SiXAa3ykkK6jnsrnWTDRfUNrFHhJLTV7sNpQ7ZiSRHiH/sQ==","repository":{"url":"git+https://github.com/AntonIXO/perplexity-web-api-mcp.git","type":"git"},"_npmVersion":"10.8.2","description":"MCP server exposing Perplexity AI search, research, and reasoning tools","directories":{},"_nodeVersion":"20.20.2","dependencies":{"axios":"^1.13.5","rimraf":"^6.1.3","detect-libc":"^2.1.2","console.table":"^0.10.0","axios-proxy-builder":"^0.1.2"},"glibcMinimum":{"major":2,"series":35},"_hasShrinkwrap":true,"devDependencies":{"prettier":"^3.8.1"},"preferUnplugged":true,"supportedPlatforms":{"x86_64-apple-darwin":{"bins":{"perplexity-web-api-mcp":"perplexity-web-api-mcp"},"zipExt":".tar.xz","artifactName":"perplexity-web-api-mcp-x86_64-apple-darwin.tar.xz"},"aarch64-apple-darwin":{"bins":{"perplexity-web-api-mcp":"perplexity-web-api-mcp"},"zipExt":".tar.xz","artifactName":"perplexity-web-api-mcp-aarch64-apple-darwin.tar.xz"},"x86_64-pc-windows-gnu":{"bins":{"perplexity-web-api-mcp":"perplexity-web-api-mcp.exe"},"zipExt":".zip","artifactName":"perplexity-web-api-mcp-x86_64-pc-windows-msvc.zip"},"x86_64-pc-windows-msvc":{"bins":{"perplexity-web-api-mcp":"perplexity-web-api-mcp.exe"},"zipExt":".zip","artifactName":"perplexity-web-api-mcp-x86_64-pc-windows-msvc.zip"},"aarch64-pc-windows-msvc":{"bins":{"perplexity-web-api-mcp":"perplexity-web-api-mcp.exe"},"zipExt":".zip","artifactName":"perplexity-web-api-mcp-x86_64-pc-windows-msvc.zip"},"x86_64-unknown-linux-gnu":{"bins":{"perplexity-web-api-mcp":"perplexity-web-api-mcp"},"zipExt":".tar.xz","artifactName":"perplexity-web-api-mcp-x86_64-unknown-linux-gnu.tar.xz"},"aarch64-unknown-linux-gnu":{"bins":{"perplexity-web-api-mcp":"perplexity-web-api-mcp"},"zipExt":".tar.xz","artifactName":"perplexity-web-api-mcp-aarch64-unknown-linux-gnu.tar.xz"}},"artifactDownloadUrls":["https://github.com/AntonIXO/perplexity-web-api-mcp/releases/download/v0.12.1"],"_npmOperationalInternal":{"tmp":"tmp/perplexity-web-api-mcp_0.12.1_1785334297213_0.5848417722129939","host":"s3://npm-registry-packages-npm-production"}},"0.12.2":{"artifactDownloadUrls":["https://github.com/AntonIXO/perplexity-web-api-mcp/releases/download/v0.12.2"],"bin":{"perplexity-web-api-mcp":"run-perplexity-web-api-mcp.js"},"dependencies":{"axios":"^1.13.5","axios-proxy-builder":"^0.1.2","console.table":"^0.10.0","detect-libc":"^2.1.2","rimraf":"^6.1.3"},"description":"MCP server exposing Perplexity AI search, research, and reasoning tools","devDependencies":{"prettier":"^3.8.1"},"engines":{"node":">=14","npm":">=6"},"glibcMinimum":{"major":2,"series":35},"keywords":["perplexity","mcp","ai","search"],"license":"MIT","name":"@antonlx/perplexity-web-api-mcp","preferUnplugged":true,"repository":{"type":"git","url":"git+https://github.com/AntonIXO/perplexity-web-api-mcp.git"},"scripts":{"fmt":"prettier --write **/*.js","fmt:check":"prettier --check **/*.js","postinstall":"node ./install.js"},"supportedPlatforms":{"aarch64-apple-darwin":{"artifactName":"perplexity-web-api-mcp-aarch64-apple-darwin.tar.xz","bins":{"perplexity-web-api-mcp":"perplexity-web-api-mcp"},"zipExt":".tar.xz"},"aarch64-pc-windows-msvc":{"artifactName":"perplexity-web-api-mcp-x86_64-pc-windows-msvc.zip","bins":{"perplexity-web-api-mcp":"perplexity-web-api-mcp.exe"},"zipExt":".zip"},"aarch64-unknown-linux-gnu":{"artifactName":"perplexity-web-api-mcp-aarch64-unknown-linux-gnu.tar.xz","bins":{"perplexity-web-api-mcp":"perplexity-web-api-mcp"},"zipExt":".tar.xz"},"x86_64-apple-darwin":{"artifactName":"perplexity-web-api-mcp-x86_64-apple-darwin.tar.xz","bins":{"perplexity-web-api-mcp":"perplexity-web-api-mcp"},"zipExt":".tar.xz"},"x86_64-pc-windows-gnu":{"artifactName":"perplexity-web-api-mcp-x86_64-pc-windows-msvc.zip","bins":{"perplexity-web-api-mcp":"perplexity-web-api-mcp.exe"},"zipExt":".zip"},"x86_64-pc-windows-msvc":{"artifactName":"perplexity-web-api-mcp-x86_64-pc-windows-msvc.zip","bins":{"perplexity-web-api-mcp":"perplexity-web-api-mcp.exe"},"zipExt":".zip"},"x86_64-unknown-linux-gnu":{"artifactName":"perplexity-web-api-mcp-x86_64-unknown-linux-gnu.tar.xz","bins":{"perplexity-web-api-mcp":"perplexity-web-api-mcp"},"zipExt":".tar.xz"}},"version":"0.12.2","volta":{"node":"18.14.1","npm":"9.5.0"},"_id":"@antonlx/perplexity-web-api-mcp@0.12.2","bugs":{"url":"https://github.com/AntonIXO/perplexity-web-api-mcp/issues"},"homepage":"https://github.com/AntonIXO/perplexity-web-api-mcp#readme","_integrity":"sha512-2gXIinmhUw1GU8BDfcdK5QGqQGM5JByIi1DkcZWHmomutyffkMg7rIfZajwre0NCsXQyxM6xqr4r6nAQWPkLpA==","_resolved":"/home/runner/work/perplexity-web-api-mcp/perplexity-web-api-mcp/npm/perplexity-web-api-mcp-npm-package.tar.gz","_from":"file:npm/perplexity-web-api-mcp-npm-package.tar.gz","_nodeVersion":"20.20.2","_npmVersion":"10.8.2","dist":{"integrity":"sha512-2gXIinmhUw1GU8BDfcdK5QGqQGM5JByIi1DkcZWHmomutyffkMg7rIfZajwre0NCsXQyxM6xqr4r6nAQWPkLpA==","shasum":"b6526db39baedde7fef6b323682ff883636b851d","tarball":"https://registry.npmjs.org/@antonlx/perplexity-web-api-mcp/-/perplexity-web-api-mcp-0.12.2.tgz","fileCount":11,"unpackedSize":57318,"signatures":[{"keyid":"SHA256:DhQ8wR5APBvFHLF/+Tc+AYvPOdTpcIDqOhxsBHRwC7U","sig":"MEYCIQCLn+igF4zQFezVQRoODBf328fL9kBwl3FZmsvGzWdgpAIhAIUk0bkyc3A//1u+O3sE4JmEdNixVpOOAIm67ptb1Ztn"}]},"_npmUser":{"name":"antonlx","email":"antonix@devpins.org"},"directories":{},"maintainers":[{"name":"antonlx","email":"antonix@devpins.org"}],"_npmOperationalInternal":{"host":"s3://npm-registry-packages-npm-production","tmp":"tmp/perplexity-web-api-mcp_0.12.2_1785764705731_0.08581224889229899"},"_hasShrinkwrap":true}},"time":{"created":"2026-07-29T14:11:36.947Z","modified":"2026-08-03T13:45:06.047Z","0.12.1":"2026-07-29T14:11:37.363Z","0.12.2":"2026-08-03T13:45:05.887Z"},"bugs":{"url":"https://github.com/AntonIXO/perplexity-web-api-mcp/issues"},"license":"MIT","homepage":"https://github.com/AntonIXO/perplexity-web-api-mcp#readme","keywords":["perplexity","mcp","ai","search"],"repository":{"type":"git","url":"git+https://github.com/AntonIXO/perplexity-web-api-mcp.git"},"description":"MCP server exposing Perplexity AI search, research, and reasoning tools","maintainers":[{"name":"antonlx","email":"antonix@devpins.org"}],"readme":"# Perplexity Web API MCP Server\n\n

\n \n \"Install\n \n \n \"Install\n \n \n \n \n

\n\nMCP (Model Context Protocol) server that exposes Perplexity AI search, research, and reasoning capabilities as tools.\n\nThis repository is the [`AntonIXO` fork](https://github.com/AntonIXO/perplexity-web-api-mcp) of the [original project by `mishamyrt`](https://github.com/mishamyrt/perplexity-web-api-mcp). Fork releases are published as [`@antonlx/perplexity-web-api-mcp`](https://www.npmjs.com/package/@antonlx/perplexity-web-api-mcp).\n\n## No API Key Required\n\nThis MCP server uses your Perplexity account session directly — **no API key needed**.\n\nPerplexity offers a separate [paid API](https://docs.perplexity.ai/guides/pricing) with per-request pricing that is charged independently from your Pro subscription. With this MCP, you don't need to pay for API access — your existing Perplexity subscription (or even a free account) is enough.\n\nSimply extract the session token from your browser cookies, and you're ready to use Perplexity search, research, and reasoning in your IDE.\n\n## Tokenless Mode\n\nThe server can run **without any authentication tokens**. In this mode:\n\n- Only `perplexity_search` (links only) and `perplexity_ask` (answer with sources) are available — `perplexity_research` and `perplexity_reason` require tokens.\n- Both tools use the `turbo` model; `PERPLEXITY_ASK_MODEL` and `PERPLEXITY_REASON_MODEL` cannot be set (the server will throw an error if they are).\n- File attachments (`files` parameter) are unavailable — they require tokens.\n\nTo use tokenless mode, simply omit `PERPLEXITY_SESSION_TOKEN` from your configuration.\n\nFor full access to all tools and model selection, provide your session token as described in the [Configuration](#configuration) section below.\n\n## Requirements\n\n### Supported Platforms\n\n- macOS (arm64, x86_64)\n- Linux (x86_64, aarch64)\n- Windows (x86_64)\n\n## Installation\n\nFor MCP clients, use the npm package directly (no global install required):\n\n```bash\nnpx -y @antonlx/perplexity-web-api-mcp\n```\n\nTo install the latest prebuilt binary on macOS or Linux:\n\n```bash\ncurl --proto '=https' --tlsv1.2 -LsSf https://github.com/AntonIXO/perplexity-web-api-mcp/releases/latest/download/perplexity-web-api-mcp-installer.sh | sh\n```\n\nOn Windows PowerShell:\n\n```powershell\npowershell -ExecutionPolicy Bypass -c \"irm https://github.com/AntonIXO/perplexity-web-api-mcp/releases/latest/download/perplexity-web-api-mcp-installer.ps1 | iex\"\n```\n\nPlatform archives and checksums are available on the [GitHub Releases page](https://github.com/AntonIXO/perplexity-web-api-mcp/releases/latest).\n\n## Configuration\n\n### Getting Your Token\n\nThis server requires a Perplexity AI account. You need to extract the session token from your browser cookies:\n\n1. Log in to [perplexity.ai](https://www.perplexity.ai) in your browser\n2. Open Developer Tools (F12 or right-click → Inspect)\n3. Go to Application → Cookies → `https://www.perplexity.ai`\n4. Copy the value of `__Secure-next-auth.session-token` → use as `PERPLEXITY_SESSION_TOKEN`\n\n### Environment Variables\n\n- `PERPLEXITY_SESSION_TOKEN` (optional): Perplexity session token (`__Secure-next-auth.session-token` cookie). Required for `perplexity_research`, `perplexity_reason`, `perplexity_computer`, `perplexity_view_chat`, connectors, and file attachments. The CSRF token is fetched automatically — no `PERPLEXITY_CSRF_TOKEN` needed.\n- `PERPLEXITY_ASK_MODEL` (optional, requires token): Model for `perplexity_ask`.\n Valid values: `turbo`, `pro-auto` (default), `pro-upgraded`, `sonar`, `nemotron-3-super`, `claude-4.6-sonnet`, `claude-4.6-opus`, `gemini-3.0-flash`, `gemini-3.0-pro`, `gpt-5-pro`, `gpt-5.3-codex`, `gpt-5.4`, `gpt-5.4-mini`, `gpt-5.2`, `gpt-5.2-pro`, `grok-4.1`.\n- `PERPLEXITY_REASON_MODEL` (optional, requires token): Model for `perplexity_reason`.\n Valid values: `gemini-3.1-pro` (default), `gemini-3.0-flash-high`, `claude-4.6-sonnet-thinking`, `claude-4.6-opus-thinking`, `gpt-5-thinking`, `gpt-5.1-thinking`, `gpt-5.2-thinking`, `gpt-5.4-thinking`, `grok-4.1-reasoning`, `kimi-k2.5-thinking`.\n- `PERPLEXITY_COMPUTER_MODEL` (optional, requires token): Model for `perplexity_computer`.\n Valid values: `asi`, `asi-beta`, `claude-4.6-sonnet` / `claude-4.6-sonnet-thinking`, `claude-4.6-opus` / `claude-4.6-opus-thinking` (default), `gpt-5.4`, `kimi`, `qwen`.\n- **`raw:` escape hatch:** any of the three model vars also accepts `raw:` to pass an arbitrary Perplexity preference string straight through, for models newer than this build's validated list (e.g. `PERPLEXITY_REASON_MODEL=raw:glm_5_2`). No recompile needed. **Important:** each mode only accepts models from its own family — `PERPLEXITY_ASK_MODEL` takes an \"ask\"-family preference, `PERPLEXITY_REASON_MODEL` takes a \"reasoning\"-family preference, and the two are not interchangeable. Setting a reasoning-only model (e.g. GLM 5.2 via `raw:glm_5_2`) as `PERPLEXITY_ASK_MODEL` will make `perplexity_ask` silently return `answer: null` instead of erroring — Perplexity's backend accepts the (wrong-family) preference string but produces no output for it. If `perplexity_ask` or `perplexity_reason` starts returning `null` answers after changing a model var, check that the model actually belongs to that tool's family before assuming a rate-limit or auth issue. See [Discovering available models](#discovering-available-models) below for how to find which family a model preference belongs to.\n- `PERPLEXITY_TIMEOUT_SECS` (optional, default: `30`): Request timeout in seconds for fast modes (search, ask, reason).\n- `PERPLEXITY_LONG_TIMEOUT_SECS` (optional, default: `600`): Request timeout in seconds for long-running modes — **Deep Research**, Computer, and Document Review. Raise this if deep-research runs are being cut off.\n **Note:** this timeout lives inside the Perplexity HTTP client only. If you're also being cut off by your *MCP client's own* tool-call timeout (a separate, client-side setting — see [Progress heartbeats](#progress-heartbeats-for-long-running-tools) below), raising this variable alone won't help; you also need to raise (or auto-extend via progress) the client's timeout.\n- `PERPLEXITY_PROGRESS_INTERVAL_SECS` (optional, default: `10`): How often (in seconds) to emit `notifications/progress` heartbeats during a tool call, when the caller's request included a `progressToken`. See [Progress heartbeats](#progress-heartbeats-for-long-running-tools) below.\n- `PERPLEXITY_INCOGNITO` (optional, default: `true`): Whether requests should use Perplexity's incognito mode.\n Valid values: `true` or `false`\n\n### Discovering Available Models\n\nPerplexity doesn't publish a stable model list in this server's code alone — the typed enums (`SearchModel`, `ReasonModel`, `ComputerModel` in `crates/perplexity-web-api/src/models.rs`) are a **snapshot** that goes stale as Perplexity ships new models. Two live sources let you check what's actually available on your account before using the `raw:` escape hatch:\n\n1. **Model config endpoint** (no login required): [`https://www.perplexity.ai/rest/models/config`](https://www.perplexity.ai/rest/models/config) — returns the full JSON list of every model Perplexity's frontend knows about, grouped by which mode(s) it's valid for (`ask`, `reason`/copilot, `computer`/agentic, etc.) along with its internal `preference` string (the exact value you pass to `raw:`). This is the authoritative source for \"does model X exist\" and \"which family (ask vs reason vs computer) does it belong to\" — cross-check against this before filing a bug report about a model returning `null`.\n2. **Rate limit / usage endpoint** (requires a logged-in browser session — Cloudflare blocks non-browser requests): [`https://www.perplexity.ai/rest/rate-limit/all`](https://www.perplexity.ai/rest/rate-limit/all) — returns your account's current quota status per feature (`remaining_pro`, `remaining_research`, `remaining_labs`, `remaining_agentic_research`, plus per-source monthly limits). Also exposed as the `perplexity_usage` MCP tool, but that tool call hits the same Cloudflare wall as any other non-browser client and will return an HTTP 403 unless you're proxying through a real browser session (see the code comment on `perplexity_usage` in `crates/perplexity-web-api-mcp/src/server.rs` for the current status of that limitation).\n\nTo pull the config JSON from a terminal:\n\n```bash\ncurl -s 'https://www.perplexity.ai/rest/models/config' | jq .\n```\n\nIf you find a model preference that isn't in the typed enums yet, either (a) use it immediately via `raw:` — no recompile needed — or (b) open a PR adding it to `models.rs` so it gets schema validation and shows up in tool descriptions.\n\n### Progress Heartbeats for Long-Running Tools\n\n`perplexity_research`, `perplexity_computer`, and `perplexity_document_review` (and, less commonly, `perplexity_reason` on a heavy query) can legitimately hold Perplexity's SSE connection open for **minutes** with no intermediate bytes — this looks identical to a hung request from an MCP client's point of view.\n\nPer the [MCP Lifecycle spec's Timeouts section](https://modelcontextprotocol.io/specification/2025-06-18/basic/lifecycle#timeouts), a client **MAY** reset its own per-request timeout clock every time it receives a `notifications/progress` message tied to that request's `progressToken` — but the spec doesn't let a *server* force this; it's entirely a client-side opt-in (`resetTimeoutOnProgress` in the TypeScript SDK, or equivalent in other clients). A `SHOULD`-level absolute maximum timeout still applies on top, regardless of progress notifications, so this cannot make a request run forever.\n\nThis server implements the server side of that contract:\n\n- If (and only if) the caller's `tools/call` request includes `_meta.progressToken`, a background task sends a `notifications/progress` message every `PERPLEXITY_PROGRESS_INTERVAL_SECS` (default 10s) for the duration of the call.\n- `progress` is a monotonically increasing tick counter — Perplexity's streaming search API doesn't expose a real percent-complete signal, and the spec only requires `progress` to increase, not to mean anything specific. `message` explains this in plain English for clients that surface it to a human (\"Still working — Perplexity request in flight, no timeout yet.\").\n- The heartbeat is a `tokio` task tied to the tool call's lifetime: it's spawned right before the underlying HTTP request and aborted on drop, whichever way the call ends (success, error, or client cancellation).\n- No `progressToken` in the request → **zero overhead**, no task is spawned.\n\n**This does not replace `PERPLEXITY_LONG_TIMEOUT_SECS`** — that variable still governs how long this server's own HTTP client will wait for Perplexity's API before giving up. The two settings solve different halves of the same problem: `PERPLEXITY_LONG_TIMEOUT_SECS` controls how long *this server* is willing to wait, and the progress heartbeat controls how long *your MCP client* is willing to wait, provided your client supports `resetTimeoutOnProgress` (or equivalent) and actually sends a `progressToken`. Many MCP clients (including some simple stdio wrappers) don't set a `progressToken` at all, in which case only the client's static configured timeout applies and you'll need to raise that directly in your client's config.\n\n### Claude Code\n\n```bash\nclaude mcp add perplexity --env PERPLEXITY_SESSION_TOKEN=\"your-session-token\" -- npx -y @antonlx/perplexity-web-api-mcp\n```\n\n### Cursor, Claude Desktop & Windsurf\n\nI recommend using the one-click install badge at the top of this README for Cursor.\n\nFor manual setup, all these clients use the same `mcpServers` format:\n\n| Client | Config File |\n|--------|-------------|\n| Cursor | `~/.cursor/mcp.json` |\n| Claude Desktop | `claude_desktop_config.json` |\n| Windsurf | `~/.codeium/windsurf/mcp_config.json` |\n\n```json\n{\n \"mcpServers\": {\n \"perplexity\": {\n \"command\": \"npx\",\n \"args\": [\"-y\", \"@antonlx/perplexity-web-api-mcp\"],\n \"env\": {\n \"PERPLEXITY_SESSION_TOKEN\": \"your-session-token\"\n }\n }\n }\n}\n```\n\n### Zed\n\nAdd following following to `context_servers` in your [settings file](https://zed.dev/docs/configuring-zed.html#settings-files):\n\n```json\n{\n \"context_servers\": {\n \"perplexity\": {\n \"command\": \"npx\",\n \"args\": [\"-y\", \"@antonlx/perplexity-web-api-mcp\"],\n \"env\": {\n \"PERPLEXITY_SESSION_TOKEN\": \"your-session-token\"\n }\n }\n }\n}\n```\n\n### VS Code\n\nI recommend using the one-click install badge at the top of this README for VS Code, or for manual setup, add to `.vscode/mcp.json`:\n\n```json\n{\n \"servers\": {\n \"perplexity\": {\n \"type\": \"stdio\",\n \"command\": \"npx\",\n \"args\": [\"-y\", \"@antonlx/perplexity-web-api-mcp\"],\n \"env\": {\n \"PERPLEXITY_SESSION_TOKEN\": \"your-session-token\"\n }\n }\n }\n}\n```\n\n### Codex\n\n```bash\ncodex mcp add perplexity --env PERPLEXITY_SESSION_TOKEN=\"your-session-token\" -- npx -y @antonlx/perplexity-web-api-mcp\n```\n\n### Building from Source\n\nSource build instructions, including optional cargo features, are documented in [CONTRIBUTING.md](CONTRIBUTING.md).\n\n### Other MCP Clients\n\nMost clients can be manually configured to use the `mcpServers` wrapper in their configuration file (like Cursor). If your client doesn't work, check its documentation for the correct wrapper format.\n\n## Docker\n\nBuild the fork's image locally:\n\n```bash\ndocker build -t perplexity-web-api-mcp .\n\ndocker run -d \\\n -p 8080:8080 \\\n -e PERPLEXITY_SESSION_TOKEN=\"your-session-token\" \\\n perplexity-web-api-mcp\n```\n\nThe container exposes the MCP server via Streamable HTTP at `http://localhost:8080/mcp`.\nThe Docker image is built with `--features streamable-http`; local/source builds need the same feature if you want HTTP transport.\n\nConfigure your MCP client to connect:\n\n```json\n{\n \"mcpServers\": {\n \"perplexity\": {\n \"url\": \"http://localhost:8080/mcp\"\n }\n }\n}\n```\n\n### Environment Variables (Docker-specific)\n\n| Variable | Default | Description |\n|----------|---------|-------------|\n| `MCP_TRANSPORT` | `streamable-http` | Transport mode. `stdio` or `streamable-http` (requires the `streamable-http` cargo feature) |\n| `MCP_HOST` | `0.0.0.0` | Host address to bind |\n| `MCP_PORT` | `8080` | Port to listen on |\n\nThe [authentication token, model variables, and incognito flag](#configuration) described above work the same way in Docker.\n\n## Available Tools\n\n### `perplexity_search`\n\nQuick web search using the `turbo` model. Returns only links, titles, and snippets — no generated answer.\n\n**Best for:** Finding relevant URLs and sources quickly.\n\n**Parameters:**\n\n- `query` (required): The search query or question\n- `sources` (optional): Array of sources — `\"web\"`, `\"scholar\"`, `\"social\"`, plus connectors like\n `\"google_drive\"`, `\"gcal\"` / `\"google_calendar\"`, `\"github_mcp_direct\"` / `\"github\"`,\n `\"hugging_face\"` / `\"huggingface\"`, `\"notion_mcp\"` / `\"notion\"`, `\"linear_alt\"` / `\"linear\"`,\n `\"slack_direct\"` / `\"slack\"`, `\"jira_mcp_merge\"` / `\"jira\"`, `\"confluence_mcp_merge\"` / `\"confluence\"`,\n `\"microsoft_teams_mcp_merge\"` / `\"teams\"`, `\"onedrive\"`, `\"sharepoint\"`, `\"dropbox\"`, `\"box\"`.\n Unknown connector IDs are passed through unchanged. Defaults to `[\"web\"]`\n- `language` (optional): Language code, e.g., `\"en-US\"`. Defaults to `\"en-US\"`\n\n> File attachments are not supported by this tool.\n\n### `perplexity_ask`\n\nAsk Perplexity AI a question and get a comprehensive answer with source citations. By default uses the best model (Pro auto mode) when authenticated, or `turbo` in tokenless mode. Can be configured via `PERPLEXITY_ASK_MODEL`.\n\n**Best for:** Getting detailed answers to questions with web context.\n\n**Parameters:** Same as `perplexity_search`, plus:\n\n- `files` (optional, requires token): Array of file attachments for document analysis. See [File Attachments](#file-attachments).\n\n### `perplexity_reason`\n\nAdvanced reasoning and problem-solving. By default uses Perplexity's `sonar-reasoning` model, but can be configured via `PERPLEXITY_REASON_MODEL`.\n\n**Best for:** Logical problems, complex analysis, decision-making, and tasks requiring step-by-step reasoning.\n\n**Parameters:** Same as `perplexity_ask`.\n\n### `perplexity_research`\n\nDeep, comprehensive research using Perplexity's sonar-deep-research (`pplx_alpha`) model.\n\n**Best for:** Complex topics requiring detailed investigation, comprehensive reports, and in-depth analysis. Provides thorough analysis with citations.\n\n**Parameters:** Same as `perplexity_ask`.\n\n### `perplexity_view_chat`\n\nLoads the context of an existing Perplexity search/chat and asks Perplexity to extract or summarize it.\n\n**Parameters:**\n\n- `url_or_id` (required): Full `perplexity.ai/search/...` URL, thread UUID, or thread slug\n- `query` (optional): A specific question about the thread. By default, returns a detailed question-by-question breakdown\n\nRequires `PERPLEXITY_SESSION_TOKEN`. Internally this uses Perplexity's follow-up mechanism, so the inspection query may appear as a new turn in the referenced thread.\n\n## File Attachments\n\n`perplexity_ask`, `perplexity_research`, and `perplexity_reason` accept an optional `files` parameter for document analysis. **Requires authentication token.**\n\nEach entry in the `files` array must have:\n\n- `filename` (required): Filename with extension, e.g. `\"report.pdf\"` or `\"notes.txt\"`\n- `text` (mutually exclusive with `data`): Plain-text file content. Use for `.txt`, `.md`, `.csv`, `.json`, source code, etc.\n- `data` (mutually exclusive with `text`): Base64-encoded binary content. Use for `.pdf`, `.docx`, images, etc.\n\n**Example — plain text:**\n\n```json\n{\n \"query\": \"Summarise the key points\",\n \"files\": [\n {\n \"filename\": \"notes.txt\",\n \"text\": \"Meeting notes: Q1 revenue up 12%...\"\n }\n ]\n}\n```\n\n**Example — binary file (PDF):**\n\n```json\n{\n \"query\": \"What does this contract say about termination?\",\n \"files\": [\n {\n \"filename\": \"contract.pdf\",\n \"data\": \"JVBERi0xLjQK...\"\n }\n ]\n}\n```\n\nMultiple files can be passed in a single request — they are uploaded to Perplexity's storage in parallel before the query is sent.\n\n## Response Format\n\n`perplexity_search` returns only web results:\n\n```json\n{\n \"web_results\": [\n {\n \"name\": \"Source name\",\n \"url\": \"https://example.com\",\n \"snippet\": \"Source snippet\"\n }\n ]\n}\n```\n\n`perplexity_ask`, `perplexity_research`, and `perplexity_reason` return a full response:\n\n```json\n{\n \"answer\": \"The generated answer text...\",\n \"web_results\": [\n {\n \"name\": \"Source name\",\n \"url\": \"https://example.com\",\n \"snippet\": \"Source snippet\"\n }\n ],\n \"follow_up\": {\n \"backend_uuid\": \"uuid-for-follow-up-queries\",\n \"attachments\": []\n }\n}\n```\n\n## License\n\nMIT\n","readmeFilename":"README.md"}