Portkey

We just raised our $15M Series A to scale our unified control plane for production AI. AI is now mission-critical infrastructure, and we’re building the reliability layer so it never breaks. Thanks to our investors @ElevCap + @lightspeedvp portkey.ai/blog/series-a-…

We processed 1 Trillion llm tokens and 1 Million llm spend today on Portkey at least this saas is thriving

Super to see so many Lightspeed portfolio companies presenting here at @MontySummit in LA - @ScrutAutomation, @acceldataio, @composio, @PixxelSpace, Bridgetown Research, @PortkeyAI - all AI-native approaches to GRC, data management, tools automation, AI gateway, consulting / research and more. @awaisahmedna @jumbld @AayushGhoshCho2 @KaranVaidya6 @hershsah

@chamath You should use @PortkeyAI for model switching and team based token cost budgeting. @jumbld

@adityamishra @jumbld Thank you! Checking asap

@PortkeyAI Your email verification seems to be down @jumbld

Shipped something today that sounds boring but is genuinely the most requested feature we've had at @PortkeyAI One tab. Every AI provider, model, API key, broken down by cost, latency, and success rate. When you're routing 22M+ requests across Anthropic, OpenAI, and Bedrock every day... you just want the damn summary.

Great to see @jumbld and the team at @PortkeyAI turning consistent customer feedback into a shipping moment: a unified view across providers, models, API keys, and workspaces. Small product decisions like this make a big difference for platform teams. This is what customer-centricity looks like 👏

@honeyhive missed this one? join us here -> luma.com/yy93e6e2

Room full of agent builders, everyone sharing what they're actually building this was fun @HoneyHive!

1. The Missing Control Layer 🧠 Most teams track AI costs in spreadsheets and pray they do not get 502 errors at 2am. @PortkeyAI fixes this. They built the ultimate control panel that sits exactly between your app and your LLM providers. It handles routing, observability, and guardrails from one clean interface.

That video is from @PortkeyAI and the product lives up to the creative. @jumbld have been building the infrastructure layer for LLM apps.  One API, one control panel, 1,600+ providers. OpenAI, Anthropic, Google, open-source.  All in one place. portkey.ai

Taking to engineers who are at scale, looking at millions of traces and multiple subagents trajectories and hand-off happening in context on challenges posed on UX, filtering, traceability of these. Plus evals against failure in hand-off and tool failures. Had tough time keeping up with @rachittshah pace of dropping insights and my ability to process it. 😅 Good un-conference style but free flowing conversation event from @PortkeyAI

@dkhare 🚀🚀

Amazing to see traffic (and customers!) ramp on @PortkeyAI. The last six months have been 🔥 ! portkey.ai/rankings/

@shobhitic this is @nxrxnxn 's brainchild. also drop by next time!

saw this cool LED outside of portkey's office that shows their token usage leaderboard for different models

What’s the best way to compare LLM outputs? We collaborated with @arizeai to see what rigorous LLM output comparison actually requires: • Evaluations that run offline (against datasets) and online (against live traffic) • Versioned, reusable evaluators • Explanations so failures are diagnosable • Traces that attach routing, prompt version, and model context to every scored call When evals attach directly to traces, you can see which model change, prompt revision, or routing decision caused it, and validate fixes before rollout. Full breakdown here -> portkey.ai/blog/the-best-…

@InsomniacPM22 @LassoSecurity agreed!

@PortkeyAI @LassoSecurity Spot on! and the attack surface scales with how many tool definitions you stuff into context. Most agents ship 50+ tool schemas when they only need 3-4 per turn. Smarter tool selection before the LLM call reduces both latency and exposure.

MCP gives your AI agents real power with access to database and internal systems. But every tool call is a new attack surface. See how Portkey's MCP Gateway and @LassoSecurity work together to secure AI Agents - portkey.ai/blog/securing-…

Sometimes your guardrails have dependencies. You might want to: → Run a PII detection check first, and then run compliance validation on the sanitized output → Check the request format before running an expensive LLM-based content safety check With the new `sequential` flag on Portkey, checks run one after another in the order you define, and if an earlier check fails, the rest don't execute. See how you can set up sequential guardrails →#configuring-guardrail-actions" target="_blank" rel="nofollow noopener">portkey.ai/docs/product/g…

@JackQuinn @ElevCap @lightspeedvp yess!!

This is a meaningful step for production AI. Observability, routing, guardrails, and cost tracking are becoming baseline requirements once agents move from demos into real workflows. A unified control layer across providers is a smart way to reduce fragility. What gets especially interesting from here is how those traces evolve into full decision reconstruction — not just seeing what was called, but understanding how authority, context, and policy state shaped the outcome over time. Reliability is the foundation. Governed autonomy is what compounds on top of it.

We just raised our $15M Series A to scale our unified control plane for production AI. AI is now mission-critical infrastructure, and we’re building the reliability layer so it never breaks. Thanks to our investors @ElevCap + @lightspeedvp portkey.ai/blog/series-a-…

@TheIshanGoswami @jumbld @ElevCap @lightspeedvp thank you! super grateful for your partnership!

@jumbld @ElevCap @lightspeedvp Congrats a ton guys!! We use portkey here at exa

Huge news: our team at Portkey has raised $15M in Series A funding led by @ElevCap w/ @lightspeedvp 🎉 As agentic workflows become a reality and businesses recognize AI as mission-critical infrastructure, we’re here to ensure it never breaks. (1/5)