{"slug": "praxist-autonomous-r-d-system-with-parallel-research-agents", "title": "Praxist – Autonomous R&D system with parallel research agents", "summary": "Sapient Inc. released Praxist, an autonomous research system that coordinates parallel research agents, task-owned evaluation, durable evidence, and generation-to-generation synthesis for measurable, computer-executable research. The system integrates with OpenAI's Codex and supports open-source model APIs, with installation via a one-line command and a setup wizard covering licensing, privacy, and runtime profiles.", "body_md": "Praxist is an autonomous research system for measurable, computer-executable research. It coordinates parallel research peers, task-owned evaluation, durable evidence, and generation-to-generation synthesis.\n\nPraxist treats research as a persistent process rather than a sequence of disconnected prompts. Use it when a project already runs and its objective is measurable, but the best path forward is still unknown.\n\nInstall the complete runtime integrations and finish first-use setup with one command:\n\n```\npython3 -m pip install --index-url https://pypi.org/simple \"praxist[agents,codex]\" && praxist setup --interactive --install-skills codex\n```\n\nThe local wizard covers the Fair Source License, User Agreement, privacy,\nruntime profile, masked credentials, Codex skills, writable examples,\nand readiness checks. It does not select a research project or launch a run.\nFor Claude Code, use the\n[host-specific one-line command](/sapientinc/PRAXIST/blob/main/docs/getting-started/installation.md#install-and-configure).\n\nFor an agent-managed installation, open Codex:\n\n```\ncodex --yolo\n```\n\nThen ask it to install and configure Praxist using the packaged OOBE runbook, and to stop after readiness checks.\n\nBefore starting research, read the [Quickstart](/sapientinc/PRAXIST/blob/main/docs/getting-started/quickstart.md)\nand [Your First Task](/sapientinc/PRAXIST/blob/main/docs/getting-started/first-task.md). They describe the\nseparate takeover step and the project contract it creates.\n\nChoose **Codex-native mode** to use an existing Codex subscription without an\nAPI key. For sustained research, Praxist generally favors\n[open-source model APIs](/sapientinc/PRAXIST/blob/main/docs/guides/open-source-model-apis.md) with a high\nobserved cache-hit rate. The setup wizard also supports other API-backed\nprofiles.\n\nWe recommend Codex as the interface for operating Praxist. Praxist is not a replacement for Codex: Codex remains the interactive agent that understands your project, communicates with you, and uses development tools. Praxist adds the persistent research loop, parallel peers, evidence protocols, scheduling, and lifecycle control.\n\nAfter installation, open Codex in the root of an already runnable research\nproject and invoke `$praxist-takeover`\n\n. The takeover skill inspects readiness,\ncreates or repairs the task harness, validates its evaluator and evidence\ncontract, and launches the run after the required gates pass. A precise brief\nproduces a better research plan; include the objective, metrics, constraints,\nresources, exploration choices, and whether launch is authorized.\n\n## Example takeover brief\n\n```\n$praxist-takeover\n\nTreat the current directory as the existing runnable research project. Verify\nthe baseline and its evaluation path before changing anything.\n\nOptimize <primary metric and direction> while preserving <key constraints>.\nUse <peer count> peers for up to <generation count> generations within\n<time or cost budget>. Use the runtime and model provider selected during\nsetup. <Allow or disable> literature search, <enable or disable> QD, and\n<enable or disable> generation-zero DIG.\n\nDo not download new datasets or replace required project assets. Build a\nseparate task harness with explicit metric directions, baseline provenance,\nprotocol-integrity checks, evidence maturity rules, and justified retention\nlanes. After readiness checks pass, <launch immediately in detached mode / ask\nfor confirmation>. Report the task path, run ID, evidence contract, generation\nclose policy, and monitor command.\n```\n\nOther bundled skills:\n\n| Skill | Purpose |\n|---|---|\n`praxist-takeover-codex` |\nNo-key takeover using the saved Codex login |\n`praxist-onboarding` |\nExplain Praxist and inspect local readiness |\n`praxist-task-initialization` |\nBuild or repair a task harness without launching |\n`praxist-interactive-task-init` |\nDesign a task through confirmation-first setup |\n`praxist-control` |\nStart, stop, resume, monitor, and inspect runs |\n`praxist-diagnostic` |\nDiagnose run health and produce reports |\n`praxist-scientific-research` |\nGather sourced literature and benchmark context |\n`praxist-runtime-install` |\nInstall or repair runtime dependencies and credentials |\n`terminal-line-plot` |\nDraw metric trends in the terminal |\n\nSee [Agent Skills](/sapientinc/PRAXIST/blob/main/docs/user-guide/skills.md) for invocation syntax and the\ngenerated [Skills Reference](/sapientinc/PRAXIST/blob/main/docs/reference/skills.md) for the complete\ncontracts.\n\n| Capability | Purpose |\n|---|---|\n| Parallel research peers | Explore competing hypotheses and implementations concurrently |\n| Multi-generation synthesis | Carry useful evidence and strategy into later generations |\n| Durable evidence lanes | Preserve candidates through incubator, frontier, and Gems state |\n| Multi-metric evaluation | Rank task-defined evidence, including Pareto-optimal tradeoffs |\n|\n\n| Praxist owns | The task project owns |\n|---|---|\n| Research orchestration, lifecycle, evidence protocols, replay, scheduling, and extension interfaces | Research objective, executable code, evaluator, metrics, baselines, prompts, roles, and domain constraints |\n\nPraxist contains no task-specific scientific assumptions. A task remains the single source of truth for what should be tested and what counts as valid evidence.\n\n```\npraxist status --json\npraxist --monitor --latest\npraxist stop <run_id>\npraxist resume <run_dir>\n```\n\n`Ctrl-C`\n\ncloses only the monitor; it does not stop the research run.\n\n```\npraxist examples list\npraxist examples install rocket_booster_recovery\npraxist examples install rocket_booster_recovery_rust\n```\n\nComplete examples are writable reference projects. `templates/tasks/`\n\ncontains\nreplaceable scaffolding for building new task harnesses. The two Rocket Booster\nRecovery examples demonstrate the same research problem through Python/JAX and\nnative Rust implementations.\n\n| Status | Requirement |\n|---|---|\n| Required | CPython 3.11+ |\n| Required to launch research | A runnable project with measurable evaluation |\n| Required for skill-driven operation | Codex or Claude Code; direct CLI operation remains available without either |\n| Authentication: choose one | A saved Codex login for Codex-native mode, or a supported provider API key |\n| Continuously release-tested | Linux on CPython 3.11 and 3.12 |\n| Compatibility target | macOS and other CPython 3.11+ environments; run `praxist doctor` before research |\n\nTask-specific datasets and compute dependencies remain owned by the task project.\n\nSee the [platform support matrix](/sapientinc/PRAXIST/blob/main/docs/operations/platform-support.md) for the\ndifference between release-qualified hosts and compatibility targets.\n\nRead the ** Praxist documentation**\nor open it with:\n\n```\npraxist docs\n```\n\nNo local documentation server is required.\n\nRead the [contribution guide](/sapientinc/PRAXIST/blob/main/.github/CONTRIBUTING.md) together with the\n[Code of Conduct](/sapientinc/PRAXIST/blob/main/.github/CODE_OF_CONDUCT.md) before participating. Source\nmaintainers should also follow the architecture and maintenance contract in\n[AGENTS.md](/sapientinc/PRAXIST/blob/main/AGENTS.md). See the [Privacy Notice](/sapientinc/PRAXIST/blob/main/docs/legal/PRIVACY.md) for\nPraxist's data-handling terms.\n\nContact: [praxist@sapient.inc](mailto:praxist@sapient.inc)\n\n## Show questions and answers\n\nPraxist is an autonomous research system for measurable research problems that can be executed on a computer. It turns an already runnable project into a continuous, evidence-driven research run.\n\nAcross successive generations, parallel research agents develop candidate solutions; evaluators convert results into structured evidence; and a planning panel synthesizes that evidence into the research agenda for the next generation. The cycle continues until the search converges or the budget is exhausted.\n\nYou provide a runnable project and a measurable objective. Praxist orchestrates the research process that searches for the best-performing solution.\n\nAutoML tunes parameters within a predefined search space. Praxist runs the full research loop.\n\nParallel research agents can change methods, architectures, and strategies.\nEvidence from evaluation shapes the agenda for the next generation, while the\n[Deep Innovation Gate (DIG)](/sapientinc/PRAXIST/blob/main/docs/guides/deep-innovation-gate.md) and\n[Quality-Diversity (QD)](/sapientinc/PRAXIST/blob/main/docs/guides/qdig-cohort-allocator.md) allocation help\nthe system escape local optima.\n\nPraxist is closer to a self-directing research team than a search tool. If your researchers are already iterating on a problem manually, Praxist takes over the iteration loop itself.\n\nPraxist delivers the most value when three conditions are met:\n\n**The objective is measurable:** there is at least one metric that meaningfully distinguishes better from worse, with a clear optimization direction.**The project already runs:** the baseline code, environment, and required data or simulator are in place and work without Praxist.**The best path forward is unknown.**\n\nIf a prerequisite is missing, Praxist stops and tells you exactly what is needed. It will not silently download unspecified datasets, invent a simulator, or fabricate baseline performance. That is a deliberate design principle.\n\nNo API key is required in Codex-native mode; Praxist uses your authenticated Codex session. We also recommend using your own API key to access supported model APIs.\n\nAPI costs are set by the provider and vary by model and usage. Total cost also depends on parallelism, the number of generations, and evaluation runtime. For cost-sensitive runs, start with a small representative workload before scaling up.\n\nPraxist provides three layers of protection:\n\n**Project isolation:** Praxist does not modify your original project. Run artifacts are stored separately.**Credentials:** API keys are entered through a masked local prompt and are not exposed in commands, shell history, or conversations.**Data collection:** Praxist does not collect data used in your experiments. It collects only limited system-level operational information, which you can disable at any time.\n\nPraxist uses three safeguards:\n\n**Preregistration:** Metrics, evaluation protocols, baselines, and acceptance thresholds are defined before the run.**Consistent evaluation:** Every candidate is measured through the same evaluator, and invalid or suspicious results are excluded.**End-to-end provenance:** Every reported improvement includes the evidence and lineage needed to inspect and reproduce it.\n\nWe recommend reviewing what the selected solution changed and testing it again in your own environment. Praxist's results are designed to be verifiable, and your own validation should be the final test.\n\nPraxist does not guarantee a specific metric improvement. It provides a rigorous research process and auditable evidence.\n\nIf a run does not meet its target, you still receive a negative-result evidence package, an audit report, and recommendations on whether to stop or redirect the research.\n\nA negative result can still be valuable: it rules out tested approaches with evidence and helps prevent further investment in an unproductive direction.\n\nPraxist is licensed under the Fair Source License Agreement 1.0. The precise description is source-available: the complete source code is publicly available and may be viewed, downloaded, and modified. Subject to the license terms, Praxist may be used for internal business purposes and deployed within your own organization.\n\nOrganizations with aggregate annual revenue, including revenue from affiliates, below US$1 million may use Praxist commercially at no charge. Once annual revenue reaches or exceeds that threshold, the organization must contact the Licensor, Sapient Intelligence Pte Ltd, to negotiate a Commercial License.\n\nThe revenue threshold does not apply to qualifying teaching and academic research conducted by institutions of higher education, public research institutions, and nonprofit academic research organizations.\n\n**Generated outputs:** no attribution is required for internal use. If an\noutput is published externally or otherwise made available to third parties,\nthe product-name attribution \"Praxist by Sapient Intelligence\" must be\nretained.\n\nThis FAQ is a summary only. If it conflicts with the Fair Source License Agreement 1.0, the terms of the license agreement control.\n\nIf you use Praxist in your research, please cite:\n\n```\n@misc{li2026praxistexperimentalartifactssolution,\n      title={Praxist: From Experimental Artifacts to Solution Lineages},\n      author={Jin Li and Ahmed Murtadha and Zhiyu Wang and Qiwen Chen and William Chen and Yifei Wu and Guan Wang and Andy L. Siy and Jiayi Yang and Mengsha Huang and Wenhao Li and Yixuan Liu and Shuailin Pan and Mingli Yuan and Sen Song and Yuhao Sun},\n      year={2026},\n      eprint={2608.25955},\n      archivePrefix={arXiv},\n      primaryClass={cs.MA},\n      url={https://arxiv.org/abs/2608.25955},\n}\n```\n\n", "url": "https://wpnews.pro/news/praxist-autonomous-r-d-system-with-parallel-research-agents", "canonical_source": "https://github.com/sapientinc/PRAXIST", "published_at": "2026-08-28 02:00:21+00:00", "updated_at": "2026-08-28 02:17:58.207933+00:00", "lang": "en", "topics": ["ai-research", "ai-agents", "ai-tools", "developer-tools"], "entities": ["Sapient Inc.", "Praxist", "Codex", "OpenAI"], "alternates": {"html": "https://wpnews.pro/news/praxist-autonomous-r-d-system-with-parallel-research-agents", "markdown": "https://wpnews.pro/news/praxist-autonomous-r-d-system-with-parallel-research-agents.md", "text": "https://wpnews.pro/news/praxist-autonomous-r-d-system-with-parallel-research-agents.txt", "jsonld": "https://wpnews.pro/news/praxist-autonomous-r-d-system-with-parallel-research-agents.jsonld"}}