aiXplain SDK
Build, govern, deploy, and monitor single- or multi-agent systems from Python.
- Source repo
- aixplain/aiXplain
- Stars
- ★ 59
- Last updated
- 7d ago
- License
- Apache-2.0
- Primary language
- Python
- FA score
- 29/100 · Major gaps
At a glance
- How it runs
- Works with
- Universal · cross-platform
- Cost
- Free tier plus a paid hosted plan
- Setup effort
- Low · running in minutes
- You'll need
- Typical use
- A Python developer adding a research assistant that can call web search and return concise, source-grounded findings.
- Not a fit if
- Teams that need fully offline runs; run metadata is sent via ipinfo.io
- Users unwilling to create an aiXplain account and API key
- Teams still on the deprecated SDK v1 factory API
- Source review
- 29/100 · Major gaps 8 safety controls not found
What does this agent do, and when should you use it?
aiXplain SDK is a Python-facing runtime and deployment platform covering the agent lifecycle from construction and evaluation through deployment, monitoring, and evolution. It supports knowledge retrieval, data tasks, custom logic, integrations, and multi-agent teams while exposing a catalog of more than 900 models, tools, and integrations through one entry point. Runtime controls include asset permissions, action allow-lists, rate and usage limits, output inspection, and step-level traces. The project says the same agent definition can run in cloud, on-premises, edge, or local environments, although the supplied evidence does not include installation details or a compatibility matrix for those targets.
The SDK reads AIXPLAIN_API_KEY, retrieves marketplace models or tools, and defines agents with instructions, tools, subagents, and Inspectors. During a run, a planner divides the goal, an orchestrator routes work, subagents and tools perform operations, and Inspectors validate or edit results under runtime policy. agent.save() promotes a definition to a persistent, versioned endpoint, while agent.run() executes a query and returns a typed result whose output is available at result.data.output. The supplied material also documents RAG-style knowledge agents, MCP asset endpoints, execution traces, and cloud, on-premises, edge, and local delivery claims.
- A Python developer adding a research assistant that can call web search and return concise, source-grounded findings.
- An engineering team coordinating research, analysis, or custom-logic subagents beneath one supervising agent and output policy.
- An enterprise that needs action allow-lists, per-asset permissions, rate limits, usage limits, and access control around agent execution.
- A platform team seeking to preserve one agent definition across cloud, on-premises, edge, and local environments.
- A developer using an MCP-compatible IDE or coding agent who wants marketplace assets exposed through hosted MCP endpoints.
How do you install or deploy this agent?
Run the documented commands in a Python environment with shell access:
pip install aixplain
export AIXPLAIN_API_KEY=<your-key>The supplied material does not specify a minimum Python version, supported operating systems, an offline installation procedure, or additional dependencies for the claimed local and on-premises deployments.
How do you use this agent?
Using the documented v2 API, create the client, retrieve and restrict a tool, define the agent, save it, and run a query:
from aixplain import Aixplainaix = Aixplain()
search_tool = aix.Tool.get("tavily/tavily-web-search/tavily")
search_tool.allowed_actions = ["search"]
agent = aix.Agent(name="Research agent",
description="Answers questions with concise web-grounded findings.",
instructions="Use the search tool when needed and cite key findings.",
tools=[search_tool],
)
agent.save()
result = agent.run(query="Who is the CEO of OpenAI? Answer in one sentence.")
print(result.data.output)For a team, pass agents through subagents, policies through inspectors, and call agent.save(save_subcomponents=True) to save the components together. The README mentions API, CLI, and MCP access, but it does not supply a general CLI workflow or complete API deployment procedure.
What are this agent's strengths and limitations?
- One SDK covers agent definition, execution, persistent deployment, runtime governance, and step-level tracing.
- The documented execution model includes planning, orchestration, subagents, tools, and output inspection.
- A single key provides access to more than 900 listed models, tools, and integrations, reducing per-provider integration work.
- Governance controls explicitly include allow-lists, asset permissions, rate and usage limits, and access control.
- The project documents MCP support and claims that definitions can run across cloud, on-premises, edge, and local environments.
- The quick start requires an aiXplain account, API key, and network access; cloud resources use consumption-based or subscription pricing.
- The supplied evidence does not identify supported Python versions, operating systems, complete self-hosting steps, or feature differences between deployment modes.
- No test suite, benchmarks, availability commitments, or asset-by-asset compatibility evidence is included in the supplied material.
- The Inspector sample uses an intentionally inert predicate, so it does not demonstrate behavior under real policy violations, tool failures, or timeouts.
- SDK v1 support ends on August 1, 2026, requiring legacy users to move to v2.
How does this agent compare with similar options?
Compared with assembling separate provider SDKs, authentication, rate limiting, queues, autoscaling, governance, and tracing, aiXplain offers a unified SDK, marketplace, and runtime. The tradeoff is that the quick start depends on an aiXplain API key and cloud assets remain subject to its pricing and service availability. Although the README claims local and infrastructure-independent operation, the supplied evidence is not detailed enough to compare deployment tradeoffs.
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| aiXplain SDK This agent | 29 · Major gaps | Library / SDKFreemium | ★ 59 | 7d ago | Python | — |
| AgenticX Multi-Agent Platform | 64 · Some gaps | CLIFree + model costs | ★ 977 | 1d ago | Python | Claude Code · OpenAI API · Claude API |
| AgentScope 2.0 | 70 · Some gaps | Library / SDKFree + model costs | ★ 35k | 2d ago | Python | OpenAI API · Claude API |
| Swarms Multi-Agent Framework | Insufficient evidence | Library / SDKFree + model costs | ★ 7.4k | today | Python | Claude Code · OpenAI API · Claude API |
How does FollowAgents rate this agent?
Why each dimension lost points
Evidence shows: no implementation or documentation of permission management, user confirmation, data flow transparency, sensitive data handling, dependency security, external effects, rollback, or source attribution. All trust criteria are absent, hence score 0.
Evidence shows: test suite and CI configuration indicate some consistency; dependencies are listed in pyproject.toml but not pinned; error handling exists in tests but no user-friendly failure messages. Thus self-consistency, dependency availability, and failure messages each score 1.
Evidence shows: README describes target audience (Python programmers) and scenarios (building AI agents); capability boundaries not clearly defined; trigger precision not defined; environment fit (Python versions, OS) stated in pyproject.toml. Thus audience and scenarios score 2, capability boundaries and trigger precision each score 1, environment fit scores 2.
Evidence shows: README and documentation links exist but information architecture incomplete; install notes mentioned in README but not detailed; naming stability not explicit; examples and FAQ limited; known limitations not stated; license is Apache-2.0; version number exists but no changelog; maintenance responsibility not clear. Thus information architecture scores 2, install notes, naming stability, examples and FAQ, versioning each score 1, known limitations scores 0, license scores 2, maintenance responsibility scores 1.
Evidence shows: output usability not explicitly stated; marginal value indicated in README (900+ models); cost-benefit not discussed. Thus output usability scores 1, marginal value scores 2, cost-benefit scores 1.
Evidence shows: claims in README not supported by evidence; cross-source verification limited; facts and inferences not separated. Thus claim traceability scores 1, cross-source corroboration scores 1, fact-inference separation scores 0.
- Not found in source: least-privilege scopingGrant only what the task needs: a dedicated account or read-only token, scoped to specific directories and repos.
- Not found in source: confirmation before actingTurn on (or add) a confirmation step before it acts, and try it in a sandbox or test environment before real data.
- Not found in source: data-flow disclosureWatch which external services it contacts (proxy or firewall logs) and keep sensitive data out until you know where it goes.
- Not found in source: sensitive-data handlingUse dedicated, low-privilege, revocable API keys — never production credentials — and keep secrets out of logs.
- Not found in source: dependency securityPin versions and run a dependency audit (npm audit, pip-audit) before installing; prefer running it in a container.
- Not found in source: disclosed external effectsEstablish which external systems it writes to, sends to or changes, and verify with test accounts or repos before production.
- Not found in source: rollback or recovery pathBack up first, or work on a git branch or snapshot, so its changes can be undone.
- Not found in source: verifiable attributionInstall from the official repo or registry and check the publisher and URL to avoid look-alike packages.
- No permission management or user confirmation mechanisms, may execute arbitrary code or access sensitive data.
- Dependencies not pinned, supply chain risk.
- Known limitations not stated, potential hidden defects.
- Publisher identity unverified, evaluate with caution.
Recent releases
Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.
-
v0.2.48 0.2.48
Release notes excerpt
Release Notes [0.2.48](https://github.com/aixplain/aiXplain/compare/0.2.47...v0.2.48) (2026-09-11) Everything merged since 0.2.47. This is the first tag to use the v prefix (v0.2.48), which the release pipeline requires. Critical fix - packaging: a source build made outside a git checkout shipped 3 of 169 modules with no error and no warning. [tool.setuptools.packages.find].include was a bare "aixplain", which fnmatches the full dotted package name and so discarded all 26 subpackages; the missing setuptools-scm file-finder had been masking it. Now "aixplain*", wi …View on GitHub ↗ -
Release notes excerpt
Release Notes [0.2.47](https://github.com/aixplain/aiXplain/compare/0.2.44...0.2.47) (2026-07-22) Combined changelog for everything since 0.2.44. This release includes and supersedes the 0.2.45rc1 pre-release and more. Critical fix - v2-agent: Agent.get() + save() no longer silently overwrites the agent's LLM with the default (GPT-5.4). On 0.2.43/0.2.44 the fetched agent never decoded the model field from the API response, so agent.llm fell back to DEFAULT_LLM and any subsequent save() persisted it — resetting the agent's configured LLM (#922). …View on GitHub ↗ -
Release notes excerpt
Release Notes [0.2.44](https://github.com/aixplain/aiXplain/compare/0.2.43...0.2.44) (2026-04-07) Features - v2-actions: Add unified Actions -> Action -> Inputs -> Input hierarchy for models and tools - Models now expose model.actions["run"].inputs with model.inputs as a shorthand - Tools now expose tool.actions["action_name"].inputs for action discovery and parameter configuration - Inputs support schema-aware validation, defaults, reset behavior, and dot/bracket access patterns - rlm: Add Recursive Language Model support to the SDK …View on GitHub ↗